Performance View User s Guide

Transcription

1 Titlepage Performance View User s Guide Document 3509 SPECTRUM Management

2 Copyright Notice Document Copyright present Aprisma Management Technologies, Inc., 273 Corporate Drive, Portsmouth, NH USA. All rights reserved worldwide. Use, duplication, or disclosure by the United States government is subject to the restrictions set forth in DFARS (c)(1)(ii) and FAR Liability Disclaimer Aprisma Management Technologies, Inc. ("Aprisma") reserves the right to make changes in specifications and other information contained in this document without prior notice. In all cases, the reader should contact Aprisma to inquire if any changes have been made. The hardware, firmware, or software described in this manual is subject to change without notice. IN NO EVENT SHALL APRISMA, ITS EMPLOYEES, OFFICERS, DIRECTORS, AGENTS, OR AFFILIATES BE LIABLE FOR ANY INCIDENTAL, INDIRECT, SPECIAL, OR CONSEQUENTIAL DAMAGES WHATSOEVER (INCLUDING BUT NOT LIMITED TO LOST PROFITS) ARISING OUT OF OR RELATED TO THIS MANUAL OR THE INFORMATION CONTAINED IN IT, EVEN IF APRISMA HAS BEEN ADVISED OF, HAS KNOWN, OR SHOULD HAVE KNOWN, THE POSSIBILITY OF SUCH DAMAGES. Trademark, Service Mark, and Logo Information SPECTRUM, IMT, and the SPECTRUM IMT/VNM logo are registered trademarks of Aprisma Management Technologies, Inc., or its affiliates. APRISMA, APRISMA MANAGEMENT TECHNOLOGIES, the APRISMA MANAGEMENT TECHNOLOGIES logo, MANAGE WHAT MATTERS, DCM, VNM, SpectroGRAPH, SpectroSERVER, Inductive Modeling Technology, Device Communications Manager, SPECTRUM Security Manager, and Virtual Network Machine are unregistered trademarks of Aprisma Management Technologies, Inc., or its affiliates. For a complete list of Aprisma trademarks, service marks, and trade names, go to All referenced trademarks, service marks, and trade names identified in this document, whether registered or unregistered, are the intellectual property of their respective owners. No rights are granted by Aprisma Management Technologies, Inc., to use such marks, whether by implication, estoppel, or otherwise. If you have comments or concerns about trademark or copyright references, please send an to spectrum-docs@aprisma.com; we will do our best to help. Performance View User s Guide Page 2

3 Restricted Rights Notice (Applicable to licenses to the United States government only.) This software and/or user documentation is/are provided with RESTRICTED AND LIMITED RIGHTS. Use, duplication, or disclosure by the government is subject to restrictions as set forth in FAR (June 1987) Alternate III (g)(3) (June 1987), FAR (June 1987), or DFARS (c)(1)(ii) (June 1988), and/or in similar or successor clauses in the FAR or DFARS, or in the DOD or NASA FAR Supplement, as applicable. Contractor/manufacturer is Aprisma Management Technologies, Inc., 121 Technology Drive, Durham, NH In the event the government seeks to obtain the software pursuant to standard commercial practice, this software agreement, instead of the noted regulatory clauses, shall control the terms of the government's license. Virus Disclaimer Aprisma makes no representations or warranties to the effect that the licensed software is virusfree. Aprisma has tested its software with current virus-checking technologies. However, because no anti-virus system is 100 percent effective, we strongly recommend that you write-protect the licensed software and verify (with an anti-virus system in which you have confidence) that the licensed software, prior to installation, is virus-free. Contact Information Aprisma Management Technologies, Inc. 273 Corporate Drive Portsmouth, NH Phone: U.S. toll-free: Web site: Performance View User s Guide Page 3

4 Contents Preface 8 Who Should Read This Guide...8 Using This Manual...8 Questions about SPECTRUM Documentation...9 Overview 10 Monitoring System Components...10 Analyzing Performance View Data...11 Optimizing Performance...12 SpectroSERVER Tuning...12 Determining Your Network Needs...12 How To s 14 How To Start Performance View...14 How To Evaluate SPECTRUM Performance...15 Thread Latency...15 Threads and Thread Latency...15 Types of Threads that Affect Performance...16 Thread Information View...17 Poll/Log Tabbed Page...18 Average Poll Latency Graph...18 Poll Threads In Use Graph...19 Average Log Latency Graph...19 Performance Considerations...20 Log Threads In Use Graph...20 Timer/Notify Tabbed Page...20 Theory of Operations...21 Average Timer Latency Graph...22 Performance Considerations...22 Timer Threads In Use Graph...23 Average Notification Latency Graph...23 Notification Threads in Use Graph...23 Swapping/Paging...24 Performance View Page 4

5 Contents Contents Indicators of Memory Problems...24 Physical Memory Utilization - Main Tabbed Page...25 Disk Utilization Graph - Disk Tabbed Page...25 Queue Length Graph - Disk Tabbed Page...26 Network I/O Graph - Network Tabbed Page...26 Paging Activity Graph - Memory Tabbed Page...26 Too Many Applications Running...26 Memory Tabbed Page...27 CPU Tabbed Page...27 Performance Considerations...27 Too Many Client Connections...28 How To Tune SpectroSERVER Performance...29 Polling Intervals...29 Default Polling and Logging Intervals...29 Staggering Polling Intervals to Reduce SpectroSERVER Workload...30 Using the Global Attribute Editor...31 Accessing the Global Attribute Editor...32 Using the Command Line Interface...32 Setting Polling Intervals...32 Setting Polling Intervals for Applications...33 Disabling Polling for Endpoints...34 Using the Model Information View...35 Disabling Polling on a Device...35 Modifying Polling on a Device...35 Real-Time Views and Performance...35 Modifying Icon Poll Time...36 Icon Graph Control...36 How To Determine the Number of SpectroSERVERs Needed...37 Sample SPECTRUM Sizing...37 Obtaining Sizing Data...40 How To Use Health Report...41 Running Health Report from the Main Window...43 Starting Data Collection...43 Stopping Data Collection...44 Saving a Health Report...46 Opening a Health Report...47 Running Health Report from the Command Line...48 Performance View Page 5

6 Contents Contents Interpreting Health Reports...49 SpectroSERVER Capacity...51 Subsystem Data...52 Analysis...52 Analysis Narratives...53 How To Set Preferences...59 Displaying the Preferences Window...60 Toggling Save Settings...60 Toggling Confirm Exit...60 Toggling Show Warnings at Exit...60 Toggling Show Server List at Startup...61 Toggling Report When Complete...61 Specifying Addresses...61 Customizing the Title Font...61 Customizing the Label Font...62 Customizing Chart Line Colors...62 Choose a Color Window...62 Windows 67 Main Window...67 Menu Bar...69 Menu Descriptions...69 File Menu...69 View Menu...70 Options Menu...71 Tool Bar...72 Tabbed Pages...72 Pop-Up Menu...74 Status Bar...75 Main Tabbed Page...76 CPU Tabbed Page...78 Memory Tabbed Page...79 Disk Tabbed Page...80 Network Tabbed Page...81 Poll/Log Tabbed Page...82 Health Report Tabbed Page...83 Timer/Notify Tabbed Page...84 Performance View Page 6

7 Contents Contents Preferences...85 Select SpectroSERVER...87 Connection Status...88 Index 89 Performance View Page 7

8 Preface This section describes the content and layout of this document, identifies the intended audience, and explains assumptions about prerequisite knowledge readers must have to effectively use the information herein. Welcome to the SPECTRUM Performance View User s Guide. This document explains how to start and use the Performance View application to monitor and tune the performance of the SpectroSERVER component(s) of your SPECTRUM system. Performance View is included with all SPECTRUM installation packages. If your SPECTRUM system has been installed in accordance with the SPECTRUM Installation Guide, you will be able to access Performance View either through SpectroGRAPH or from the command line. For detailed instructions, see How To Start Performance View on Page 14. Who Should Read This Guide This guide is intended for network administrators and technicians responsible for day-to-day administration of a network. This guide presumes you are familiar with the terms and concepts associated with traditional network devices such as hubs, routers, and bridges as well as with the terms and concepts associated with SPECTRUM. It also presumes that you are familiar with traditional networking models and standards i.e., the Open Systems Interconnection (OSI) model and the 802.1d, 802.3, and standards. Using This Manual This manual is divided into three main sections: Overview - This section describes SPECTRUM s Performance View application in general and discusses the performance-related concepts it was designed to address. Performance View Page 8

9 Preface Questions about SPECTRUM Documentation How To s - This section describes procedures for starting the Performance View application and using it to identify, evaluate, and troubleshoot performance problems. Procedures for customizing the Performance View user interface are also covered. Windows - This section describes the functionality of Performance View s user interface, including detailed information about performance indicators that can be monitored and administrative tasks that can be launched. Questions about SPECTRUM Documentation your questions, comments, or suggestions regarding SPECTRUM documentation to spectrum-docs@aprisma.com Performance View Page 9

10 Overview This section describes SPECTRUM s Performance View application in general and discusses the performance-related concepts it was designed to address. In SPECTRUM most network and device models offer generic Performance views that display statistics on traffic, errors, etc. (For more information, see the SPECTRUM Views document.) For the VNM model that represents SpectroSERVER, however, there is a special, more robust, Performance View application that lets you monitor both server and network performance and helps you determine how to tune your system for optimum performance. The rest of this section provides a high level look at how you can use Performance View to optimize your SPECTRUM system. The following section (How To s on Page 14) addresses the specific tasks and procedures involved, and the final section (Windows on Page 67) describes the user interface in detail. This document provides guidelines for identifying SPECTRUM performance problems and selecting the proper procedures for resolving these problems. Performance problems can be resolved either by tuning the SpectroSERVER or by adding new SpectroSERVERs to distribute the entire network load or by a combination of both methods. Performance View allows you to monitor SpectroSERVER performance and view the system resource utilization of each SpectroSERVER. Monitoring System Components Each computer system is comprised of four major components: disk, network, memory, and CPU. In order to ensure the successful operation of your SPECTRUM system, you may need to tune any or all of these components to eliminate bottlenecks. To help you detect and locate such bottlenecks, Performance View s main window (see Main Window on Page 67) includes a separate tabbed page dedicated to providing current information on each these components. There is also a Main tabbed page that lets you see both SpectroSERVER and total percentage of utilization Performance View Page 10

11 for disk, memory, and CPU at a glance. This page also displays SPECTRUM-specific information such as the number of client connections, number of models, etc. Note also that in a Distributed SpectroSERVER environment, you can switch the focus of Performance View from one server to another via the Select Server option on the main window s Options menu. In addition to the tabbed pages for monitoring overall system and network activity, there are two pages that graphically display SpectroSERVERspecific information: the Timer/Notify tabbed page, which provides information on timer and notification thread allocation and latency. the Poll/Log tabbed page, which provides information on log and poll latency problems. The log latency is the average time lapse, in seconds, between the initial logging request and the time logging occurs. Poll latency is the time lapse, in seconds, from the initialization of resource polling to the time when polling occurs. Finally, there is a Health Report tabbed page, which allows you to generate reports on the growth of the SpectroSERVER process size over any 24-hour period (see How To Use Health Report on Page 41). Analyzing Performance View Data As noted above, Performance View gives you a variety of perspectives for monitoring the performance of your SpectroSERVER. One of the tools available, the Health Report mentioned above, provides its own analysis of the data it collects. In many cases, Performance View gives a fairly direct indication of where a problem might lie. For instance, the main window s CPU tabbed page provides a real-time top ten list of processes in terms of CPU use. Exactly what constitutes a problem or bottleneck will of course depend on the particular configuration of your SPECTRUM system and on your particular network management priorities. However, throughout the How To s section of this document, an attempt has been made to provide guidelines that will help you identify problems and decide upon a course of action for resolving them. Performance View Page 11

12 Overview Optimizing Performance Optimizing Performance In general, performance problems can be resolved either by tuning the SpectroSERVER or by adding new SpectroSERVERs to distribute the entire network load or by a combination of both methods. Each method is discussed below. SpectroSERVER Tuning Once you have determined the reasons for degraded performance, you can tune the SpectroSERVER to optimize performance. This tuning is achieved primarily by modifying polling intervals and logging ratios to reduce network traffic and the resulting latency that degrades performance. Changing polling intervals or turning polling off altogether can be achieved using certain SPECTRUM applications, views, scripts or files. There are two major area where a user can tune a SpectroSERVER: Reduce load by reducing polling and logging Increase capacity by increasing memory, CPU speed, or disks For more information, refer to How To Tune SpectroSERVER Performance on Page 29. Determining Your Network Needs You will need to determine if increasing the number of SpectroSERVERs, rather than tuning, is the best means of achieving desired performance improvements. To do this, you will need to run a utility called PMCount and submit the resulting data to SPECTRUM Support or your VAR, so that the data can be entered into the SPECTRUM Online Sizing tool. The sizer will then generate a report on your network parameters detailing the estimated additional amount of generated network management traffic as a result of SPECTRUM and the number and configuration of additional SpectroSERVERs required to efficiently manage the number of models. If your sizing determines that you need to increase the number of SpectroSERVERs to two or more to manage your network efficiently, you Performance View Page 12

13 Overview Determining Your Network Needs will need to determine how best to set up a distributed network environment, which is detailed in Distributed SpectroSERVER. For more information, refer to How To Determine the Number of SpectroSERVERs Needed on Page 37. Performance View Page 13

14 This section describes procedures for starting the Performance View application and using it to identify, evaluate, and troubleshoot performance problems. Procedures for customizing the Performance View user interface are also covered. The following information is presented in terms of specific tasks and procedures you can perform with Performance View s user interface. For a component-by-component description of the interface itself, refer to Windows on Page 67. How To Start Performance View You can start Performance View from the either the VNM icon or the landscape icon in SpectroGRAPH, or you can start it from the command line. To start Performance View from the VNM icon, you can double-click in the center of the icon, or you can either highlight the icon and select View > Icon Subviews > Performance or drag the cursor over the icon and right-click to display the Icon Subviews menu. To start Performance View from the landscape icon, either highlight the icon and select View > Icon Subviews > Performance or drag the cursor over the icon and right-click to display the Icon Subviews menu. To start Performance View from the command line, navigate to your SPECTRUM installation directory and enter the following command:./pview/pview Performance View Page 14

15 How To Evaluate SPECTRUM Performance Overall SPECTRUM performance can be influenced by a variety of considerations. These considerations can be attributed to the configuration of the hardware system or to the configuration of software loaded on or interacting with the hardware system. Some general performance categories that may provide insight into a perceived degradation in SPECTRUM performance are: Thread Latency High Swapping/Paging Activity Too Many Applications Running on the SPECTRUM System Too Many Clients Connecting to the SPECTRUM System Thread Latency This section describes how to determine if thread latency is symptomatic of SPECTRUM performance problems. The types of threads and indicators of latency are discussed. The following views can help you determine if SpectroSERVER performance has degraded due to thread latency: Thread Information View (Page 17) Poll/Log Tabbed Page (Page 18) Timer/Notify Tabbed Page (Page 20) Threads and Thread Latency A thread is a a set of commands that perform a function or set of functions. Each thread is able to run independently from other threads. SpectroSERVER is a multi-threaded application. Multi-threading is a method of performing multiple tasks at the same time. While some threads are waiting (for user input, response from devices, data retrieval, etc.), other threads can be running (logging data, responding to traps, connecting to SSAPI applications, etc.). As each thread runs in the Performance View Page 15

16 How To Evaluate SPECTRUM Performance SpectroSERVER process, it takes control of the CPU for a few microseconds and then relinquishes control to allow other threads to run. SPECTRUM performance can be influenced by the timely allocation of threads. SpectroSERVER maintains a pool of threads that are shared over time: SPECTRUM s polling, logging, client request, IH timers, IH notifications, model activation, and model destruction processes. A SpectroSERVER sub-system uses threads from the pool, up to their individual limits, during periods of increased processing activity. These maximum limits prevent any one SpectroSERVER sub-system from dominating resources and consuming all of the available threads. When the common pool of threads is exhausted, new threads are created and the pool grows to meet the needs of the increase in activity. Threads that are no longer needed by a process are returned to the common pool for later use. When a thread remains unused for a specified period of time, it is removed from the pool and its resources are returned to the system. This process is called aging. Thread latency is the amount of queue time associated with the allocation of the various software threads to be used in various SPECTRUM processes (e.g., polling, logging). If SPECTRUM runs for a prolonged period of time with high thread latencies, there will be a delay in the polling of devices, logging of data, etc. and, therefore, a delay in SPECTRUM s response time. For example, a critical network device could become inoperable and there would be a delay before SPECTRUM would notify a network administrator of the problem. Thread latency is a symptom, not a cause, of SPECTRUM performance degradation. Types of Threads that Affect Performance Poll Threads Poll threads are responsible for polling devices on the network. SPECTRUM uses polling as a way to manage the operation and performance of the network. The SpectroSERVER code responsible for polling is called the Poll Manager. It is the Poll Manager that initiates polling of all of the network devices. Performance View Page 16

17 How To Evaluate SPECTRUM Performance Log Threads Log threads are responsible for logging data from the network into archive files in the SPECTRUM database. SPECTRUM uses data logging to store information about the operation and performance of the network. Timer Threads Timer threads notify inference handlers that they have registered (expressed an interest in) timers. Notification Threads Notification threads are used to notify an inference handler of a change in an attribute that it has registered for. Thread Information View The Thread Information view is not part of the panels available in the Performance view. To access the Thread Information View, highlight the VNM icon and choose View > Icon Subviews > Configuration to bring up the Landscape Configuration view. In the Advanced Information section of this view, click on the Thread Information button to launch the Thread Information view. The Thread Information view shows the threads in use, threads available, and peak value for various types of threads. It allows you to change the thread available value for each type. Generally these values should be left at their defaults. If one of these thread types is consistently running at the limit, and there are available CPU cycles, increasing a limit may reduce the associated latency. If CPU utilization is already above 80%, increasing thread limits will not increase throughput, and may actually reduce throughput due to the increased thread overhead. Poll Threads: Used to read the polled attributes for a model on the model s polling_interval. Log Threads: Used to read and log the logged attributes for a model on the model s polling_interval * poll_log_ratio. Notification Threads: Used to send notifications of attribute changes to inference handlers and Spectrum client applications. Performance View Page 17

18 How To Evaluate SPECTRUM Performance IH Timer Threads: Used to trigger timers in inference handlers. Destroy Threads: Used to send model destruction notifications to inference handlers and client applications. Model Activate Threads: Used to send model activation notifications to inference handlers and client applications. Relation Activate Threads: Used to send relation change notifications to inference handlers and client applications. Client Request Threads: Used to handle client application requests. Multi Request Threads: Used to handle multi-model requests that originate from inference handlers and client applications. Poll/Log Tabbed Page The Poll/Log Tabbed Page provides an indicator of whether the allocation of polling or logging threads is affecting SPECTRUM performance. The Poll/Log Tabbed Page has four graphs to indicate the amount of thread usage and any latency associated with polling or logging threads: Average Poll Latency Poll Threads in Use Average Log Latency Log Threads in Use Average Poll Latency Graph Poll latency is the queue time for polling resources to become available. The Average Poll Latency graph shows the average queue time for the polling process in seconds. The calculation of poll latency displayed in the graph is the interval in seconds between when a scheduled polling operation is supposed to happen and when it actually happens. For example, if the average poll latency is 10 seconds and a model is supposed to be polled every 60 seconds, then the model is actually polled every 70 seconds. Performance View Page 18

19 How To Evaluate SPECTRUM Performance The effect of running for prolonged periods with high poll latency is the delayed polling of devices. Delaying the time to poll devices can increase the response time associated with detecting a network fault. A device could go down and the system administrator would be unaware. If the average value in the Poll Latency graph is equal to or greater than 3 per second and is sustained for a considerable period of time, this indicates that SpectroSERVER performance has been degraded. Options for improving SPECTRUM performance would include: tuning the system, off-loading system demand, or upgrading the speed/capacity of the system. Poll Threads In Use Graph A poll thread is allocated to every polling operation. Poll threads are allocated from a finite number of threads that are resident in SPECTRUM. The number of poll threads that get used by the system at any one time, is proportional to the number of models to be polled on the network and the number of times each model is polled. If the number of poll threads needed exceeds the number of poll threads available, then pending poll thread requests are queued until the next poll thread becomes available. If the number of poll threads available has been exceeded, then the number of poll threads may be insufficient for the current state of the network. Call SPECTRUM Support. Average Log Latency Graph Log latency is the queue time for logging resources to become available. The Log Latency graph shows the current queue time for the logging process in seconds. The calculation of log latency as displayed in the graph is the delta in seconds between when a scheduled logging operation is supposed to happen and when it actually happens. For example, if the average log latency is 10 seconds and data is supposed to be logged every 60 seconds, then the data is actually being logged every 70 seconds. The effect of running for prolonged periods with high log latency is the delayed logging of data and could lead to other serious performance problems. Performance View Page 19

20 How To Evaluate SPECTRUM Performance Performance Considerations If the average value in the Log Latency graph is equal to or greater than 30 seconds, this indicates that SpectroSERVER performance has been degraded. Options for improving SPECTRUM performance include: tuning the system, off-loading system demand, or upgrading the speed/capacity of the system. Log Threads In Use Graph A log thread is allocated to every logging operation. Log threads are allocated from a finite number of threads resident in SPECTRUM. The number of log threads that get used by the system at any one time, is proportional to the amount of data being logged. When the number of log threads needed exceeds the number of log threads available, then pending log thread requests are queued until a log thread becomes available. If the number of log threads available has been exceeded, then the number of log threads may be insufficient for the current state of the network. Call SPECTRUM Support. Timer/Notify Tabbed Page The Timer/Notify Tabbed Page displays the number of timer or notification threads in use and any notification or timer thread latency. The Timer/Notify Tabbed Page has four graphs: Average Timer Latency Graph Timer Threads in Use Graph Average Notification Latency Graph Notification Threads in Use Graph Performance View Page 20

21 How To Evaluate SPECTRUM Performance Theory of Operations The SpectroSERVER is an event driven system. An inference handler registers for events it is interested in, and the SpectroSERVER notifies the inference handler when these events occur. Events that an inference handler can register for include model creation, model destruction, attribute value changes, association adds/removals, etc. Notification Threads A SpectroSERVER process called the Notification Manager reads attribute changes received from polled devices and then runs inference handlers to process this data. Another SpectroSERVER process called the Poll Manager is responsible for detecting changes in an attribute. When a change in an attribute value is detected, the Poll Manager alerts the Notification Manager and the Notification Manager forwards the events to each registered inference handler. Notification threads are the mechanism used to notify inference handlers of a change in an attribute. Notification threads are used to run the trig_attr_change and trig_watch_change inference handler methods. Both of these methods are notifications of attribute value changes. For example: the ifindiscards attribute has changed from a count of 110 to 150. The SpectroSERVER sends a notification thread to the inference handler that has expressed an interest in this attribute. The inference handler then runs the trig_attr_change method. Timer Threads Timer threads are used to notify inference handlers that have registered for timers (wake up calls). When an inference handler registers a "wake up call" with the SpectroSERVER, the SpectroSERVER will use a timer thread to run that inference handler's trigger method after a specified time interval. Timer threads are used to run the trig_timer inference handler method in the same way that notification threads are used to run trig_attr_change and trig_watch_change inference handler methods. For example: a router s primary address becomes non-operational and a secondary address must be used. An inference handler registers with the SpectroSERVER for a timer so it can determine when the primary address Performance View Page 21

22 How To Evaluate SPECTRUM Performance is once again operational. The inference handler then runs the trig_timer method. SpectroSERVER Figure 1: Notification and Timer Threads Inference Handler Sends Notification Thread Runs trig_attr_change trig_watch_change Sends Timer Thread Runs trig_timer Average Timer Latency Graph Timer latency is the queue time for timer threads to become available. The Timer Latency graph shows the current queue time for the allocation of timer threads in seconds. The calculation of timer latency, as displayed in the graph, is the interval in seconds between when a scheduled timer thread request is supposed to happen and when it actually happens. For example, if the average timer latency is 10 seconds and inference handler has registered for a timer thread every 60 seconds, then the inference handler s corresponding trigger method is actually being activated by the timer thread every 70 seconds. The effect of running for prolonged periods with high timer latency is the delayed activation of inference handler triggers and, therefore, network monitoring. Performance Considerations If the timer latency graph shows more than 3 seconds sustained, this indicates performance problems and you should consider tuning the SpectroSERVER. Performance View Page 22

23 How To Evaluate SPECTRUM Performance Timer Threads In Use Graph This is a rate graph displaying the number of timer threads in use. The number of timer threads that get used by the system at any one time, is proportional to the number of timer thread requests. When the number of timer threads needed exceeds the number of timer threads available, then pending timer thread requests are queued until a timer thread becomes available. If the number of timer threads available has been exceeded, then the number of timer threads may be insufficient for the current state of the network. Call SPECTRUM Support. Average Notification Latency Graph Notification latency is the queue time for notification threads to become available. The Notification Latency graph shows the current queue time for the allocation of notification threads in seconds. The calculation of notification latency, as displayed in the graph, is the interval in seconds between when a scheduled notification thread request is supposed to happen and when it actually happens. For example, if the average notification latency is 10 seconds and inference handler has registered for a notification thread every 60 seconds, then the inference handler s corresponding trigger method is actually being activated by the notification thread every 70 seconds. The effect of running for prolonged periods with high notification latency is the delayed activation of inference handler triggers and, therefore, network monitoring. Notification Threads in Use Graph The number of Notification Threads that get used by the system at any one time, is proportional to the number of attribute or watch changes occurring on the network. Once a network is up and has achieved stability, the number of Notification Threads required to monitor the system should remain steady and small. When the number of notification log threads needed exceeds the number of notification threads available, then pending notification thread requests are queued until a notification thread becomes available. If the Performance View Page 23

24 number of notification threads available has been exceeded, then the number of notification threads may be insufficient for the current state of the network. Call SPECTRUM Support. Swapping/Paging This section describes how to determine if swapping/paging activities are causing SPECTRUM performance problems. Swapping/paging is a processing technique that involves transferring data from a main storage area to an auxiliary storage area and back (i.e from memory to disk). Pages refer to the individual units of data transfer used to swap data. Considerable swapping activity indicates a shortage of system memory as data has to be temporarily transferred from memory to disk to make room for various processes to run. A large amount of paging activity is an indication of a high amount of swapping. Swap space together with the physical memory makes up the available memory or virtual memory. This is frequently a bottleneck for the overall system performance. As a rule of thumb, the system should have 2 times the amount of physical memory configured as swap space. If this space is low, reconfigure the system with more memory and swap space. The following features can help you determine if SpectroSERVER performance has degraded due to insufficient memory: Physical Memory Utilization - Main tabbed page Disk Utilization Graph - Disk tabbed page Queue Length Graph - Disk tabbed page Paging Activity Graph - Memory tabbed page SS Memory Utilization - Memory tabbed page Indicators of Memory Problems Memory management is important for achieving and maintaining SPECTRUM performance. Memory shortage has an immediate and significant effect on SpectroGRAPH fault detection response times. Performance View Page 24

25 One direct consequence of running SPECTRUM with insufficient memory is that it appears disk I/O bound. In many ways, a memory shy system can appear to be disk bound because of the high page and swap activity occurring to the disk. When memory is the primary bottleneck, only increasing the amount of memory, or decreasing the demand for memory will return the system back to acceptable levels of performance. Adding more disk capacity or speed will have a negligible effect. Indicators of a memory bottleneck are usually associated with the following: Slow response time within SpectroGRAPH Delays in fault detection High memory utilization High swapping/paging I/O Physical Memory Utilization - Main Tabbed Page The Physical Memory Utilization area of the Main Tabbed Page displays how much physical memory the system is currently using and how much memory SPECTRUM is currently using. Total physical memory shown to the right of the Used and SpectroSERVER bar graphs is the actual amount of physical memory contained in the machine to which the Performance View is connected. This value is machine dependent. If SpectroSERVER uses a large percentage of the virtual memory, you should consider upgrading the memory or allocating more swap space. Disk Utilization Graph - Disk Tabbed Page The Disk Utilization Graph records all disk transfers, including all physical disks attached to the system. Disk utilization is how busy the disk is (what percentage of time it is being used).. If the Disk Utilization Graph is high, this means that data logging might also be high. If this graph shows continuously high numbers, consider changing the logging ratio of some of your models and running the Performance View Page 25

26 Pmcount utility. See How To Determine the Number of SpectroSERVERs Needed on Page 37. Queue Length Graph - Disk Tabbed Page The Queue Length Graph displays the number of disk access requests pending. It takes a finite amount of time for a disk s heads to move to a specific location and read data from a disk or to write data to a disk. During this time, any other disk requests must wait. Normally, there should be very few disk requests in the queue. If the sustained average is over 1.0, this indicates a performance problem Network I/O Graph - Network Tabbed Page The Network I/O Graph records network I/O from the Ethernet interfaces on the system. This would include I/O activity from the SpectroSERVER. An increase can be expected when either new models are created in the database or the polling intervals are changed. Paging Activity Graph - Memory Tabbed Page The Paging Activity Graph displays the number of system pages over time. If the values of this graph are persistently high, then there is not enough physical memory. A situation like this can result from having more processes running than the available physical memory. A persistently high value indicates that the system is heavily loaded and a memory upgrade should be considered. If you see high paging activities, you should either reduce the number of non-spectrum processes running or increase the physical memory. Too Many Applications Running This section describes how to determine if there are too many applications running on the system. These applications would also include SpectroSERVER and SpectroGRAPH. The following views can help you determine if too many applications running on the system are causing performance problems: Performance View Page 26

27 Memory Tabbed Page CPU Tabbed Page Memory Tabbed Page The Memory Tabbed Page displays the top system memory users. This view can help determine which applications are consuming the most system memory. CPU Tabbed Page The CPU Tabbed Page displays the top system CPU users. This view can help determine which applications are consuming the most system CPU. Performance Considerations If SpectroSERVER CPU Utilization is low, but SPECTRUM performance is still slow, then other applications may be putting a strain on the system. Performance View Page 27

28 Too Many Client Connections This section describes how to determine if the system has too many client connections. A large number of client connections can place a heavy load on the server. There are two ways to determine the currently active client connections and the amount of requests the SpectroSERVER has received by using the following views: The Number of Clients Connected field in the Main Tabbed Page Client View Note: The number of client connections a system can support is relative to the configuration of that particular system. The Main tabbed page shows the number of clients connected at this time. If this number seems high, proceed to the Client View to find out which users are connected and how much traffic they produce. Persistently high numbers indicate that you should consider the use of multiple servers for individual users or user groups. Performance View Page 28

29 How To Tune SpectroSERVER Performance How To Tune SpectroSERVER Performance Once you have determined the reasons for degraded SPECTRUM performance, there are a variety of ways you can tune the SpectroSERVER to improve performance. These include: using Search Manager s Global Attribute Editor to enable, disable, or modify polling intervals or log ratios on specific devices; using the Model Information view to change polling intervals for endpoint devices; using Command Line Interface (CLI) scripts to modify polling intervals for a specific model or models; editing certain SPECTRUM files to disable or modify polling intervals. Polling and logging are the primary workload on a SpectroSERVER. Changes to polling and logging will have a significant impact on performance. To see the best performance, you should poll and log only what you need. After a brief overview of polling, the various methods for modifying polling intervals are described below. Polling Intervals SPECTRUM polls devices to retrieve management information. You can change the polling interval for each device. Note that the following apply: If you increase the time between polls, you will use less bandwidth for management traffic. However, you will have less frequent updates of device status. If you decrease the time between polls, you will have more frequent updates of device status. However, you will use more bandwidth for management traffic. Default Polling and Logging Intervals By default, SPECTRUM polls some devices every 60 seconds, other devices at 300 seconds, and logs statistics every 10 polling cycles. In Performance View Page 29

30 How To Tune SpectroSERVER Performance many cases polling and logging of this frequency is unnecessary and slows performance by creating network traffic and resulting latency. A good rule of thumb is to poll and log critical background devices every 60 seconds and other less critical network devices every 180 to 300 seconds. Polling/logging for endnodes, such as workstations, often can be turned off altogether to reduce network traffic and SpectroSERVER workload. Staggering Polling Intervals to Reduce SpectroSERVER Workload You can set staggered polling intervals to reduce network management traffic, spread out SpectroSERVER workload, and, at the same time, enhance fault management capabilities as shown in Figure 2 on Page 31. If all the devices in the example were using a default polling interval of 60 seconds, they would all be using SpectroSERVER resources every 60 seconds. By setting the polling interval for the router at 60 seconds and all of the other devices to 600 seconds, the SpectroSERVER resource utilization is reduced. However, you will not lose management capabilities because if anything happens to the devices downstream from the router, polling will be interrupted and an alarm will be generated. Performance View Page 30

31 Figure 2: Staggering Polling Intervals Polling Interval=60 Router WAN Link Polling Intervals=600 Polling Interval=600 Polling Interval=600 Using the Global Attribute Editor The Global Attribute Editor is an external application launched from the Search Manager Application. The Search Manager Application will search for, find and list all models of a particular criteria in the given landscape. You can then select any number of models found and perform configuration options on them. In terms of increasing SpectroSERVER performance, the following can be performed using the Global Attribute Editor: disable polling on a device or devices modification of device polling interval and log ratio Performance View Page 31

32 Accessing the Global Attribute Editor To access the Global Attribute Editor: 1 Highlight a model row or group of rows in the Search Manager Main Window. 2 Select Attributes from the Tool Bar or Management -> Set Attribute Values. Once the main window is opened, the Performance Tabbed Page is selected. You can set the following polling attributes on individual models, groups of models, or set default values on a model type. Polling Interval This selection allows you to set a new polling interval for all devices selected. The polling interval is the time interval, in seconds, at which SpectroSERVER will read all attributes of the device model that are flagged as POLLED. Polls to Log Ratio This selection allows you to set a desired log ratio for all selected devices. The number of SpectroSERVER polls of a device that occur prior to logging the attributes flagged as LOGGED into the database. The default value is 10. Using the Command Line Interface Polling may be modified or disabled using Command Line Interface (CLI) scripts found in the <SPECTRUM Installation Directory>/vnmsh directory as outlined below. Setting Polling Intervals You may individually or globally change the polling interval by model type using a Command Line Interface script. The CLI script for this purpose and general directions for its use are in the file <SPECTRUM Installation Directory> /vnmsh/sample_scripts/update_mtype. You must enter three Performance View Page 32

33 arguments for the script: the model type name, the polling interval attribute ID, and the polling interval value in seconds. 1 From the command line in the directory <SPECTRUM installation path>/vnmsh/sample_scripts, enter the following and press return: update_mtype <model type name> 0x10072 <new polling interval in seconds> Example: update_mtype Rtr_Cisco 0x At the prompt, select an individual model or all models of that type. Example: 0x <model> 0x60000 Rtr_Cisco 0x <model> 0x60000 Rtr_Cisco 0x <model> 0x60000 Rtr_Cisco Prompt:> 0x or <model> selects the individual model Prompt:> 0x60000 or Rtr_Cisco selects all models of that type 3 Press return and the system will indicate success or failure and disconnect you from CLI. In the example, the polling interval for all the selected Rtr_Cisco models will be changed to 300 seconds. Setting Polling Intervals for Applications Application model types also have polling intervals. Some application models initially have their polling interval set to zero. To set the polling interval for these application models, use the same methods described in the previous section. Tip: It is recommended that you check your inventory report for application models and run the update_mtype script on each global application model type to set a polling interval of 60 seconds. Performance View Page 33

34 Disabling Polling for Endpoints Administrators may not want to see the models of endpoints (endnodes) because of the alarms that may occur each time the endpoints, such as workstations, are powered down. Some administrators do not model endpoints for this reason. If you wish to have the endpoints modeled, but do not wish to use bandwidth with network polling traffic, you can disable polling for all models of a particular model type from the command line, as explained below, or you can disable polling on a model-by-model basis from the model s Model Information view (see Disabling Polling on a Device on Page 35). The CLI script for this purpose and general directions for its use are in the file <SPECTRUM Installation Directory>/vnmsh/sample_scripts/update_mtype. You must enter three arguments for the script: the model type name, the polling interval attribute ID, and the polling status (false). 1 From the command line in the directory <SPECTRUM installation path>/vnmsh/sample_scripts, enter the following and press return: update_mtype <model type name> 0x10071 false Example: update_mtype Host_Sun 0x10071 false 2 At the prompt, select an individual model or all models of that type. Example: 0x <model> 0x60000 Host_Sun 0x <model> 0x60000 Host_Sun 0x <model> 0x60000 Host_Sun Prompt:> 0x or <model> selects the individual model Prompt:> 0x60000 or Host_Sun selects all models of that type 3 Press return and the system will indicate success or failure and disconnect you from CLI. Performance View Page 34

35 In the example, the polling status for all the selected Host_Sun models will be disabled. Using the Model Information View Polling may be modified or disabled using the Model Information view of any individual device as outlined below. Disabling Polling on a Device You may disable polling for each device individually in the Model Information view of that model as follows: 1 Open the Model Information view for the endpoint model. 2 Set the Polling Status selector button to False. 3 Select Save All Changes from the File menu. 4 Click OK in the confirm dialog box. Modifying Polling on a Device You may alter the polling interval for each device individually in the Model Information view of that model as follows: 1 Open the Model Information view for the endpoint model. 2 Enter new polling interval (in seconds) in the Poll Interval field. 3 Select Save All Changes from the file menu. 4 Click OK in the confirm dialog box. Real-Time Views and Performance Real-time views, such as Device Topology, Chassis, Interface Device or any performance view launched against a device or an interface, are SPECTRUM views that rely on constant polling to keep them updated. Real-time views can cause significant load on the server and degrade performance. Performance View Page 35

36 In order to minimize the effect these views can have on performance, it is important to remember the following points: When not in use, close, do not minimize, real-time views. Even when minimized, these views are continually being updated; Consider reducing or disabling the polling on endpoints (such as workstations) to reduce network traffic when views like the DevTop are opened; Use the iconpollfactor resource within the appdefaults/spectrum file to reduce icon polling time for views with a large number of icons with performance graphs, e.g., a topology view with many LAN models. Modifying Icon Poll Time The iconpollfactor controls the lengthening of the polling interval for for icons with performance graphs that appear in map views (like the Topology view) in proportion to the number of these types of icons. The purpose is to reduce SpectroSERVER workload from views with large numbers of icons. The default iconpollfactor is 50. The number must be greater than 0. Decreasing the iconpollfactor increases the polling interval and vice versa. Icon Graph Control Also located in the app-defaults/spectrum file, the displaygraphonicons resource allows you to turn on icon graphs that, by default, are disabled (set to False). These LAN icon graphs poll from 4 to 7 attributes at a 10-second interval. This can be a significant load if you have a lot of LAN icons in an open view. Unless you really need these graphs, make sure that this setting is False. Performance View Page 36

37 How To Determine the Number of SpectroSERVERs Needed How To Determine the Number of SpectroSERVERs Needed The SPECTRUM Sizer determines the number of SpectroSERVERs you will need to efficiently manage your network. The sizer is used in conjunction with the PMCount utility which, when executed, provides the raw data (number of pollable models in a database, polling intervals, pollto-log ratios, number of ports, etc.) used by the sizing tool. The information on polling intervals and poll-logging ratios derived from running the PMCount utility and using that data in a subsequent sizing can help define where polling/logging can be reduced or disabled thereby improving performance. For detailed information on accessing and using the PMCount utility and on having a sizing done, contact your VAR or SPECTRUM Support (support@aprisma.com). Sample SPECTRUM Sizing Below is a sample SPECTRUM sizing report, generated by SPECTRUM Support after entering the pertinent data on a large customer network. The top three sections are summaries of the entered data. The bottom three sections (SpectroSERVER Sizing, SNMP Traffic and Please Note) contain the results of the sizing. The SpectroSERVER Sizing section is a listing of characteristics about the machines you chose to size for. Each line represents a different type of machine as well as how many are required to run SPECTRUM with the given network characteristics. The SNMP Traffic section estimates the additional amount of network management traffic generated by SPECTRUM. This does not include traffic due to real-time views, traps, or transient conditions such as model creation/reconfiguration or fault isolation. The Please Note section is a summary of any additional information you should be aware of when interpreting the sizer results. Performance View Page 37

38 How To Determine the Number of SpectroSERVERs Needed The results of a sample SPECTRUM sizing are shown below. General Sizing Information Account Traffic Control Center Sizing Note Regional NOC SPECTRUM Version 5.0 SpectroGRAPHs 1 Routers Quantity Interfaces Poll Interval (sec) Poll to Log Ratio Note Tier Tier Tier 3 Other managed devices Quantity Device Poll Interval (sec) Poll to Log Ratio Interface s Note 30 NortelPassport 160 Passport 142 Generic SNMP Generic Network Device 600 Generic SNMP Generic Network Device Newbridge 3600 Series Muxes Newbridge 2600/2700 series DTUs Performance View Page 38

39 How To Determine the Number of SpectroSERVERs Needed Quantity Device Poll Interval (sec) Poll to Log Ratio Interface s Note 130 Generic SNMP Generic Network Device Meridan SL 1 PBXs 334 Frame Relay PVC Generic SNMP Generic Network Device Bay Centillion 100s SpectroSERVER Sizing # Systems RAM (Mb) % Utilized Platform CPU Speed (MHz) OS Other Info Unix SUN Ultra Solaris PC Compaq WS 6000 Dual Pentium NT Compaq raid 0, 2 10k drives SNMP Traffic Estimated SNMP traffic load due to polling and logging: 0.80 Mbits/sec. This includes the SNMP traffic to and from the SpectroSERVER caused by normal steady-state polling and logging. It does not include SNMP traffic due to traps, transient events such as fault isolation, or any real time views that may be open. Performance View Page 39

40 How To Determine the Number of SpectroSERVERs Needed Please Note SpectroGRAPHs are assumed to be remote (i.e. reside on machines separate from the SpectroSERVERs). To have a local SpectroGRAPH, at a minimum, you must add on an additional 96 Mbytes of memory onto what the sizing has specified. This initial sizing is based on the raw SpectroSERVER capacity required to manage the network described by the information provided to Cabletron by the client. Actual requirements may vary depending on network architecture and the client s business requirements. For NT systems a striped disk configuration consisting of a minimum of two 10,000 RPM SCSI drives is required to achieve this level of performance. Obtaining Sizing Data In order for SPECTRUM Support to perform an accurate sizing on your network, you must run a utility called PMCount and provide the resulting data to SPECTRUM Support. Contact SPECTRUM Support (support@aprisma.com) for detailed information on accessing the PMCount utility. Performance View Page 40

41 How To Use Health Report How To Use Health Report Performance View s Health Report feature can be used to measure and report on the relative health of your SpectroSERVER and the system on which it is installed for any period from 6 up to 24 hours. You can start the reporting process at any time either from the main Performance View window (Page 43) or from the command line (Page 48). The command line option also lets you take advantage of your host system s native scheduling service to run the report automatically over a specified time period or at regular intervals. By default, Health Report collects CPU, disk, and memory usage data, network I/O and trap data, and poll, log, timer, and notification latency data at 10-second intervals for a 24-hour period. The full 24-hour period is recommended because it ensures the collection of a typical day s worth of data. If data representing your typical day s workload can be collected during a shorter period of time, you can, of course, use that time period instead, but be careful not to exclude data collection for jobs that run during non-business hours, such as backups executed at midnight. When data collection is complete, a health report is generated automatically (see Note below). If initiated from the Health Report tabbed page, the completed report will be displayed there (see Figure 3 on Page 42) and you can then save it to a location of your choice. If initiated from the command line, the report will be written to a file as explained on Page 48. You can also configure Health Report for automatic distribution (in plain text format) either from the command line or from the Preferences dialog accessed via the main window s Options menu (see Toggling Report When Complete on Page 61). Note: If data is not collected for the full 24 hours, a report will not be generated automatically. When you stop data collection before 24 hours have elapsed, you will be given the option of creating a health report from the data collected up to that point, as long as the minimum six hours worth of data has been collected. For more information, see Stopping Data Collection on Page 44. Performance View Page 41

42 Figure 3: How To Use Health Report Example of Completed Health Report Display The relative health of each system resource (e.g., CPU usage) is determined by analyzing the Average, Peak, and % Over Critical Value readings collected. Any values that exceed predefined thresholds are flagged (displayed in red) to indicate a potential performance problem. Performance View Page 42

43 How To Use Health Report The relative health of the SpectroSERVER is determined by applying performance algorithms to all data collected during the data collection period. If it is determined that performance has degraded, the likely causes and recommendations for improving performance will be provided in the Analysis section at the end of the report. For a more detailed discussion of report content, see Interpreting Health Reports (Page 49). Running Health Report from the Main Window From the Health Report tabbed page of Performance View s main window, you can do the following: start and stop data collection for a health report view, save, or print the current report open and print previously saved reports Detailed instructions for each of these operations are provided below. Starting Data Collection To start data collection for a health report: 1 If desired, configure Health Report for automatic distribution (see Toggling Report When Complete on Page 61). 2 In Performance View s main window, click the Health Report tab to display the Health Report page. 3 In the Data Collection panel (shown below), click the Start button. The Not Started message will be replaced briefly by Collecting Data, after which the message area will dynamically display the number of hours and minutes remaining in the default 24-hour reporting period. The amount of time remaining is also shown graphically in the progress indicator immediately above the message. Performance View Page 43

44 How To Use Health Report. When all data has been collected, Health Report will analyze the data and generate a health report, which will be displayed in the large area to the right of the Data Collection panel. As long as the new report remains on display, the average percentage of SpectroSERVER capacity used during the reporting period will also be shown in the System Capacity panel s pie-chart graph. Stopping Data Collection Data collection initiated from the Health Report tabbed page will stop automatically after data has been collected for a total of 24 hours, but you can stop (or pause) it manually at any time by clicking the Data Collection panel s Stop button. If you stop data collection and the total accumulated amount of time that data has been collected is equal to or greater than six hours, the following dialog will be displayed: Performance View Page 44

45 How To Use Health Report The three buttons in this dialog function as follows: Clicking Resume Data Collection restarts data collection from the point where you stopped it i.e., if you stopped it after one hour, it would then continue for 23 more hours until the default reporting period of 24 hours was reached. In other words, data collection time does not have to be contiguous; you can start and stop data collection for the same report as many times as you like as long as the total cumulative number of hours data is collected is equal to or greater than the minimum six-hour data collection period. Clicking Stop and Analyze Data immediately generates and displays a health report based on the data collected up to that moment. The report will remain on display until you start data collection for a new report or exit Performance View. Note that you cannot resume data collection for the same report once you have clicked this button. Clicking Stop and Delete Data ends the data collection process and deletes all data collected to that point without generating a report. If you stop Health Report before data has been collected for a total of six hours, the message will inform you that not enough data has been collected to create a report. In this case, only the Resume Data Collection and Stop and Delete Data buttons will appear in the dialog. Performance View Page 45

46 How To Use Health Report Saving a Health Report To save a new health report displayed on the Health Report tabbed page, do the following: 1 Click the Report File panel s Save button to display the standard save dialog shown below. 2 Navigate to the directory to where you want the health report saved, and then enter the file name you want it saved under. You do not need to include the (.htm) file extension as part of the file name, it will be added automatically. 3 Click OK, to save the health report to the specified location. Performance View Page 46

47 How To Use Health Report Opening a Health Report To open a previously saved Health Report, do the following: 1 Click the Report File panel s Open button to display the standard open dialog shown below. 2 Navigate to the directory containing the health report file you want to open, and then double-click on the file name. The selected Health Report will then be displayed in a separate view-only window. Note: To print a saved Health Report, open the report using your browser of choice, and then use the browser s print command to print the report. Performance View Page 47

48 How To Use Health Report Running Health Report from the Command Line You can start data collection for a health report at any time by entering the desired parameters from the command line, or you can use your system s native scheduling service to execute the command at a specified time or at regular intervals. The command-line executable is named pviewrep and is located in SPECTRUM s PView directory. The syntax for the pviewrep command is as follows: pviewrep <vnm> -c <collecttime> -e <addrlist> where vnm is the name of the SpectroSERVER to be reported on, collecttime is the number of hours for which you want to collect data, and addrlist is one or more comma-separated addresses to which you want the completed report to be sent. If you are specifying more than one address on Windows NT/2000 systems, the list must be enclosed in quotation marks (e.g., address1,address2,address3 ). Reports generated by pviewrep are saved automatically with the name of the SpectroSERVER host machine and an.htm extension. Sequential numbers are added to keep subsequent report file from overwriting existing ones. For example, the first report generated for a machine named ace will be ace.htm, the second report will be ace_1.htm, and so on. By default, pviewrep-generated reports are saved to the same directory as the last health report generated via Performance View s Health Report tabbed page. If that directory is not available, the report will be saved to the current working directory. If the file cannot be written to the current working directory, it will be written to the program s standard output stream. Note: Health reports distributed automatically via will be in plain text format rather than HTML. Performance View Page 48

49 Interpreting Health Reports How To Use Health Report The completed health report display consists of three parts. Figure 4 on Page 50 shows the first two parts: SpectroSERVER Capacity and Subsystem Data. In most cases, you will have to scroll down to view the third part, which is the Analysis section shown in Figure 5 on Page 51. Detailed description of all three areas follow the figures. Performance View Page 49

50 How To Use Health Report Figure 4: Top Section of Health Report Performance View Page 50

51 How To Use Health Report Figure 5: Bottom Section of Health Report SpectroSERVER Capacity This part of the health report provides the following information about the SpectroSERVER: Capacity Used - The average percentage of SpectroSERVER capacity being used during the data collection period. This percentage is also shown graphically in the System Capacity panel at the lower left of the Health Report tabbed page. Data Collection - This area shows the total amount of time data was collected as well as the start and end times for the individual data collection intervals that make up the total time. SpectroSERVER Process Size - Displays the amounts of RAM the SpectroSERVER was using at the beginning and end of the data Performance View Page 51

52 How To Use Health Report collection period as well as the difference (Growth) between these two amounts. Subsystem Data This part of the health report provides Average, Peak, and % Over Critical Value readings for individual parameters within the performance categories identified in the left-most column (CPU, LATENCIES, DISK, MEMORY, and NETWORK). Any Average or % Over Critical Value reading that exceeds Performance View s predetermined threshold value for that parameter is shown in red. Parameters thus flagged indicate possible performance problems and are used to determine the message displayed in the Analysis area of the report. The Average, Peak, and % Over Critical Value columns are defined as follows: Average - The sum of all the values recorded for the parameter during the whole data collection period (Total Time) divided by the number of collection points. (A collection point occurs every 10 seconds.) Peak - The highest value recorded for a parameter during the whole data collection period. % Over Critical Value - The percent of the data collection period during which a value for a parameter exceeded the predetermined threshold value for that parameter. Analysis This part of the health report (Figure 6) provides a narrative describing the results of the analysis performed on the collected data. The narratives are predetermined for each of the different scenarios described under Analysis Narratives (Page 53). Performance View Page 52

53 How To Use Health Report Figure 6: Analysis Section of Health Report Analysis Narratives The narrative that appears in the Analysis part of a health report depends on which parameter values, if any, exceed predetermined thresholds. Report results fall into three possible categories: (1) no parameters are flagged; (2) no Average reading is flagged, but one or more % Over Critical Value readings are flagged; and (3) any Average readings are flagged. The narratives that will be displayed for various scenarios under each of these categories are presented below 1 If all Average and % Over Critical Value numbers are fine (nothing was flagged red), then the following narrative will be displayed: Performance View Page 53

54 How To Use Health Report The SpectroSERVER appears to be running healthy and should be capable of handling approximately (100 - % Capacity) % more load; assuming that the type of new devices being modeled remains relatively the same, and no additional workloads are introduced (for example: high trap rates or additional SpectroWatches). 2 If all Average numbers are fine, but at least one % Over Critical Value number is flagged, then the following narrative will be displayed along with one of explanations from the subsequent bulleted list: On average, the SpectroSERVER is running within an acceptable resource utilization range, however, as indicated by a high "% over threshold" value, there are excessive periods of time where one (or more) of the system resources are over-utilized. This could be an indication that the resource could be close to a premature bottleneck. Based on the calculated values from this data collection period, it appears that the following problems might exist: (If the CPU % Over Critical Value number is flagged) There were periods of time where the CPU was too busy to handle the current workload. This could be caused by other applications running on the SpectroSERVER machine periodically consuming available CPU or that the system is approaching its maximum management capacity. Only the bare minimum SPECTRUM applications should be running on the SpectroSERVER machine. (If log latency % Over Critical Value number is flagged, but not paging) There are excessive log latencies, which could be a result of too much/too frequent logging or unexpected workloads, such as high event rates. (If paging and log latency % Over Critical Value numbers are both flagged) There was excessive memory swapping on the system. This Performance View Page 54

55 How To Use Health Report could be caused by other applications running on the SpectroSERVER machine periodically consuming available memory. Only the bare minimum SPECTRUM applications should be running on the SpectroSERVER machine. On Solaris machines, check that priority paging has been enabled. On machines running NT Server, check that the NT page stealing algorithm is set to "balance. (If paging % Over Critical Value is flagged but not log latency) There were periods of time of excessive memory swapping. Although the system is running out of memory, it is not affecting system performance yet, but could be an issue shortly. This could be caused by other applications running on the SpectroSERVER machine periodically consuming available memory. Only the bare minimum SPECTRUM applications should be running on the SpectroSERVER machine. On Solaris machines, check that priority paging has been enabled. On machines running NT Server, check that the NT page stealing algorithm is set to "balance. (If paging and disk utilization % Over Critical Value numbers are flagged, but not log latency) There were periods of time of excessive memory swapping. Although the system is running out of memory, it is not affecting system performance yet, but could be an issue shortly. This could be caused by other applications running on the SpectroSERVER machine periodically consuming available memory. Only the bare minimum SPECTRUM applications should be running on the SpectroSERVER machine. On Solaris machines, check that priority paging has been enabled. On machines running NT Server, check that the NT page stealing algorithm is set to "balance. If disk utilization % Over Critical Value is flagged but not paging) There were periods of time of excessive disk utilization. The system could be close to a disk Performance View Page 55

56 How To Use Health Report bottleneck possibly due to too much/too frequent logging or unexpected workloads, such as high event rates. If notification % Over Critical Value is flagged) It appears that there was an excessive amount of notifications being generated. This could be a result of models being created or reconfigured, or an unusually high rate of events had been generated during the data collection period. Try running the collection again. (If traps % Over Critical Value is flagged) It appears that there was an excessive amount of traps being generated. This could be caused by an unusually high rate of events being generated during the data collection period. Try running the collection again. (If only poll latency % Over Critical Value is flagged) There were periods of time of excessive polling latency. This could be a result of network problems preventing contact with a specific device(s). Check that there are no abnormal network issues and run the data collection again. (If any other scenario where all Average numbers are fine, but at least one % Over Critical Value number is flagged) There may have been a situation which caused the SpectroSERVER machine to experience abnormal behavior. Try running the data collection again, and if the problem persists, contact support. 3 If any Average value is flagged, then the following narrative will be displayed along with one of explanations from the subsequent bulleted list: The SpectroSERVER is currently running at full capacity and no other models or additional work should be introduced. Steps should be taken to minimize the current Performance View Page 56

57 How To Use Health Report workload or some of the models/work may have to be moved off to an additional SpectroSERVER. SPECTRUM support can perform a sizing to verify that your machine configuration is adequate for your existing database. (If CPU, poll, or timer latency Average is flagged) It appears that the SpectroSERVER has too much work to process. Increase polling intervals on all devices to the minimum required that will achieve business goals/needs. (If log latency and disk utilization or disk queue length Average numbers are flagged, but not paging) It appears that the SpectroSERVER is doing too much logging. Either increase the poll-log ratio (so that logging occurs less frequently) or turn off logging on models that are not necessary. A disk array running RAID 0 could help alleviate this problem if one is not already in use. (If traps Average is flagged) There are many traps/events being generated that are causing the SpectroSERVER to saturate with work. Check to see if there is a specific device that is causing a trap storm and also re-evaluate business goals/requirements to determine which traps are unnecessary and turn off trap forwarding on those devices, or turn off SPECTRUM's trap handling for those device types. (If paging or paging and disk utilization Average numbers are flagged) It appears that the system is running low on physical memory. The SpectroSERVER process should never exceed available physical RAM (in other words, the process should not be getting paged in and out as a result of a memory shortage). Any unnecessary applications should remain closed or more memory needs to be added to the Performance View Page 57

58 How To Use Health Report SpectroSERVER machine. If there is a local SpectroGRAPH running on the same system, splitting it off to run remotely would help reduce memory consumption. (If notification Average is flagged) It appears that there was an excessive amount of notifications being generated. This could be a result of models being created or reconfigured, or an unusually high rate of events had been generated during the data collection period. Try running the collection again. (If any other scenario where an Average number is flagged) There may have been a situation which caused the SpectroSERVER machine to experience abnormal behavior. Try running the data collection again, and if the problem persists, contact support. Performance View Page 58

59 How To Set Preferences How To Set Preferences Preferences lets you toggle some of Performance View s functionality and lets you customize the appearance of some of Performance View s graphical elements. You can: Toggle Save Settings at Exit - If checked, all settings you change during a Performance View session are saved and those changes are applied each subsequent time Performance View is started. Settings that are saved include: - application window size and location on screen - title and label fonts - confirm exit setting - show warnings on exit setting - visible tagged pages and their order of appearance - display server list at startup - last server connected to - last directory where health report saved to and opened from - show status bar, toolbar, tooltips - chart line colors Toggle Confirm Exit - If checked, requests confirmation before Performance View is closed. Toggle Show Warnings at Exit - If checked, shows all warnings pending such as, a Health Report has been created but not saved, before Performance View is closed. Toggle Show Server List at Startup - If checked, when Performance View is started, a list of SpectroSERVERs to which Performance View can connect is displayed. Customize the Title Font - Lets you specify the font, style, and size of graph titles. The graph title appears at the top of a graph. Customize the Label Font - Lets you specify the font, style, and size of graph labels. Labels are text elements in a graph other than the title. Performance View Page 59

60 How To Set Preferences Customize Chart Line Colors - Lets you choose the color of the line for any chart in a graph. There are twelve lines to choose from. Displaying the Preferences Window There are two ways to display the preferences window: Click the Preferences icon from the Toolbar. Select Preferences from the Options menu. Toggling Save Settings To toggle the Save Settings preference, click the check box to the left of Save Settings. When a check mark appears in the box, the preference is set. When no check mark appears in the box, the preference is unset. Toggling Confirm Exit To toggle the Confirm Exit preference, click the check box to the left of Confirm Exit. When a check mark appears in the box, the preference is set. When no check mark appears in the box, the preference is unset. Toggling Show Warnings at Exit To toggle the Show Warnings At Exit preference, click the check box to the left of Show Warnings At Exit. When a check mark appears in the box, the preference is set. When no check mark appears in the box, the preference is unset. Performance View Page 60

61 Toggling Show Server List at Startup How To Set Preferences To toggle the Show Server List At Startup preference, click the check box to the left of Show Server List At Startup. When a check mark appears in the box, the preference is set. When no check mark appears in the box, the preference is unset. Note: This preference is only applied when Performance View is started from the command line or the NT Start menu. If Performance View is started from within SpectroGRAPH, Performance View will always connect to the SpectroSERVER that SpectroGRAPH is connected to. Toggling Report When Complete To toggle the Report When Complete preference, click the check box to the left of Report When Complete. When a check mark appears in the box, the preference is set and the report will automatically be ed to the addresses you specify in the Report To field described below. When no check mark appears in the box, the preference is unset. Specifying Addresses Enter the addresses you want the report sent to in this field, separating each address with a comma. Customizing the Title Font To customize the Title font: 1 Click the Font button to display the Font Selection dialog box. 2 Choose a font, style, and size from their respective lists. As you make your selections, a sample is displayed in the Sample box. 3 Click OK to set the Title s font, style, and size. Performance View Page 61

62 Customizing the Label Font To customize the Label font: 1 Click the Font button to display the Font Selection dialog box. 2 Choose a font, style, and size from their respective lists. As you make your selections, a sample is displayed in the Sample box. 3 Click OK to set the Label s font, style, and size. Customizing Chart Line Colors To customize a Chart Line Color: 1 Each button represents the color of a chart line (attribute). This single palette of colors is shared by all charts in Performance View. Changing an attribute color will change that color in all Performance View charts. The first (left-most) color button is used to select a color for the first attribute in a chart, the second color button is used to select a color for the second attribute, etc. If you want to change the default color of any attribute, click the corresponding color button to display the Select Color window. 2 Choose a color and then click OK. You must click OK or Apply on the Preferences window before color changes will take effect. See the following section for more information on choosing a color. Choose a Color Window The Choose a Color window lets you change the color the specified chart line using any of three methods: Swatches, HSB, and RGB. The window consists of a color selection area, a Preview area, and OK, Cancel, and Reset buttons. Performance View Page 62

63 How To Set Preferences Figure 7: Select Color (Swatches) You use color selection area to select a color. The Preview area always shows the current selected color in a variety of configurations. You click OK to apply a color and click Cancel to dismiss the Color Select window without making color changes. You can click Reset at any time to set the color back to the initial color. The initial color shown is the color of the Chart Line Color button you selected from the Preferences window. Performance View Page 63

64 The way in which you select a color depends on which tab is selected; however, when a color is selected, that color becomes the default for the other methods, regardless of the method used to select the color. For example, if you choose a certain shade of red using the Swatches method, the color shown in the color selection area of the HSB and RBG methods will be that shade of red. Swatches Method To select a color using the Swatches method, click on the swatch that matches the color you want (Figure 7 on Page 63). The most recent colors selected are shown in the Recent area. HSB Method To select a color using the HSB (Hue, Saturation, Brightness) method shown in Figure 8 on Page 65, do one of the following: Performance View Page 64

65 Figure 8: Select Color (HSB Method) Move the slider up and down the color spectrum bar located in between the color square and the text boxes until you are satisfied with the hue. The color in the square changes to the reflect the hue selected. Move the small circle around in the square to select saturation and brightness. The values selected for hue, saturation, and brightness are shown in the HSB text boxes. Equivalent red, green, blue values are shown in the view-only RGB text boxes. Enter values directly into the text boxes that control hue, saturation, and brightness. Performance View Page 65

66 How To Set Preferences RGB Method To select a color using the RGB method (Figure 9 on Page 66), move each slider to adjust red, green and blue values individually. Figure 9: Select Color (RGB Method) Performance View Page 66

67 Windows Windows This section describes the functionality of Performance View s user interface, including detailed information about performance indicators that can be monitored and administrative tasks that can be launched. Performance View s user interface is basically a main window containing eight different tabbed pages. Some additional dialogs are accessible through the main window s menu bar. The main window and each of the tabbed pages and dialogs are described below. Main Window The Performance View main window (Figure 10 on Page 68) is your gateway to all Performance View information. It appears whenever you start Performance View. The main window consists of the following parts: Title Bar - Displays the name of the application (Performance View), and the name of the SpectroSERVER to which the Performance View is connected. Menu Bar - Lists the names of the available menus. Select an item from a menu to change the appearance of the Main window and launch administrative tasks. Tool Bar - Displays icons that can be selected to launch administrative tasks. Tabbed Pages - Performance specific information is presented depending on which tab is selected. Status Bar - Displays the current state of the Performance View and the location of the cursor. Performance View Page 67

68 Windows Main Window Tooltips - Provides a short description of a window element when you drag the cursor over the element and leave it there for more than one second. Pop-Up Menu - Available by clicking the right mouse button anywhere within the confines of a tabbed page. Not available from Main tabbed page. Menu selections let you customize how data collected is displayed. Figure 10: Performance View Main Window Title Bar Menu Bar Toolbar Tooltip Pop-up Menu Preferences Tabbed Page Status Bar Performance View Page 68

69 Windows Menu Bar SPECTRUM Performance View provides several menus from which you can launch Performance View tasks. Menu names appear in the menu bar across the top of the main window. When you select a menu, its corresponding pull-down menu is displayed. You open a pull-down menu by clicking on a menu title with the left mouse button. To make a selection from an open pull-down menu, click the menu item. To close a pull-down menu without making a selection, move the cursor off the menu and click the left mouse button. Menu Descriptions The File, View, and Options menus of the Performance View s main window are described individually in the following sections. File Menu Open Report - Displays a dialog that lets you locate and open a previously saved health report. This option functions in the same way as the Open button on the Health Report tabbed page (see Opening a Health Report on Page 47). Save Report - Displays a dialog that lets you save a health report in a location of your choice. This option functions in the same way as the Save button on the Health Report tabbed page (see Saving a Health Report on Page 46). The option and the button are grayed out unless there is a report waiting to be saved. Performance View Page 69

70 Windows Exit - Closes the main window and exits the Performance View application. View Menu Status Bar - Click to toggle the Status Bar On/Off. The status bar indicates the condition of the Performance View connection to the selected SpectroSERVER. Toolbar - Click to toggle the Toolbar On/Off. Refer to Tool Bar on Page 72 for more information. Tooltips - Click to toggle Tooltips On/Off. Tooltips provide a short description of a window element when you drag the cursor over the element and leave it there for more than one second. Connection Status - Click to launch the Connection Status window. For more information about the Connection Status window, refer to Connection Status on Page 88. Performance View Page 70

71 Windows CPU - Click to toggle the CPU tabbed page On/Off. Refer to CPU Tabbed Page on Page 78 for more information. Disk - Click to toggle the Disk tabbed page On/Off. Refer to Disk Tabbed Page on Page 80 for more information about the Disk tabbed page. Health Report - Click to toggle the Health Report tabbed page On/Off. Refer to Health Report Tabbed Page on Page 83 for more information. Main - Click to toggle the Main tabbed page On/Off. Refer to Main Tabbed Page on Page 76 for more information. Memory - Click to toggle the Memory tabbed page On/Off. Refer to Memory Tabbed Page on Page 79 for more information. Network - Click to toggle the Network tabbed page On/Off. Refer to Network Tabbed Page on Page 81 for more information. Poll/Log - Click to toggle the Poll/Log tabbed page On/Off. Refer to Poll/Log Tabbed Page on Page 82 for more information. Timer/Notify - Click to toggle the Timer/Notify tabbed page On/Off. Refer to Timer/Notify Tabbed Page on Page 84 for more information. Options Menu Preferences - Click to launch the Preferences window. Preferences lets you set startup, exit, font, and color options for the Performance View. Refer to Preferences on Page 85 for more information. Select Server - Click to launch the Select Server window. Select Server lets you choose a SpectroSERVER to connect to from a list of all available servers. Refer to Select SpectroSERVER on Page 87 for more information about selecting a SpectroSERVER. Performance View Page 71

72 Windows Tool Bar SPECTRUM Performance View provides several icons from which you can launch Performance View tasks. Click an icon to launch the associated task. Opens the Select SpectroSERVER window. Opens the Preferences window. Opens the Choose Report File to Save window. Opens the Choose Report File to Open window. Tabbed Pages Tabbed pages provide detailed information about a particular aspect of performance depending on the page selected. Information is presented in bar graph, line graph, and text formats. Bar graphs are used in the Main tabbed page. These graphs display data collected in ten minute running averages. Line graphs are used in the CPU, Memory, Disk, Network, Poll/Log, and Timer/Notify tabbed pages. These graphs display data collected over 60 minutes. If data is collected for more than 60 minutes, only the most recent 60 minutes worth of data is displayed. A high water line tool tip shows the exact high water value. For single line graphs, instantaneous values are shown in a text box in the top right corner of the graph. For multiple line graphs, instantaneous values are shown in color coded labels, each label representing the instantaneous value of its associated line. Performance View Page 72

73 Windows Line graphs also provide high-level marks for each line being graphed. High level marks represent the highest data value collected since data collection was last started. High level marks are located to the right of the line with which they are associated. You can reset high level marks. Line graphs also include a graph pop-up menu, which lets you customize the graph s parameters. Refer to Pop-Up Menu on Page 74. Multiple CPUs and disk drives can be graphed in the same graph. In a bar graph, labels identify which CPU or disk drive is being graphed. In a line graph, colored lines are used to identify which CPU or disk drive is being graphed. Refer to Preferences on Page 85 for information about how to customize the colors used in a line graph. Performance View Page 73

74 Windows Pop-Up Menu The pop-up menu consists of the following items: Auto Scale, Fit All Data, Reset Y Axis, Set Y Axis, Reset High Marks, Horizontal Guidelines, Vertical Guidelines. To display the menu, click the right mouse button anywhere within a line graph. Auto Scale - Toggles the Auto Scale feature on or off. Auto Scale automatically sizes the vertical axis value of a graph to the highest data value collected. For instance, if the current vertical access value is 20 and data is collected for a value of 300, the vertical axis is automatically set to 300. To toggle Auto Scale, click on the Auto Scale menu item. Auto Scale is on when a check mark is displayed to the left of the menu item. Auto Scale overrides Fit All Data and Set Y Axis. Fit All Data - Selects a Y axis scale so that all current data collected can be displayed on the graph. Fit All data overrides Set Y Axis. Reset Y Axis - Sets the Y axis to the default values. Set Y Axis - Displays the Set Y Axis dialog box. Performance View Page 74

75 Windows Max Value - Enter the maximum Y axis value. Min Value - Enter the minimum Y axis value. Divisions - Select the number of equal intervals desired between the minimum and maximum Y axis values. Click OK to set the Y axis values or Cancel to return to the current tabbed page without changing Y axis values. Reset High Marks - Each line graph has an associated high mark which represents the highest data value collected since data collection was last started. Reset High Marks sets all high marks in a graph back to zero. High marks are then determined by new data that is collected from that point forward. Horizontal Guidelines - Toggle horizontal guidelines On/Off. To toggle the horizontal guidelines, click the Horizontal Guidelines menu item. Horizontal Guidelines are on when a check mark is displayed to the left of the menu item. Vertical Guidelines - Toggle vertical guidelines on or off. To toggle the vertical guidelines, click the Vertical Guidelines menu item. Vertical Guidelines are on when a check mark is displayed to the left of the menu item. Status Bar Collecting Data... Displays the current state of the Performance View or the current status of the Health Report View. The color of the connection icon indicates the status of the connection to the designated SpectroSERVER as follows: Performance View Page 75

76 Windows Main Tabbed Page Color Status Main Tabbed Page Green Yellow Red Normal Using Backup SpectroSERVER Contact Lost The Main tabbed page consists of the Physical Memory Utilization graph, the CPU Utilization graph, the Disk Utilization graph, and General Data information. Performance View Page 76

77 Windows Physical Memory Utilization - Displays the total physical memory being used (Used), and the amount of physical memory being used by the SpectroSERVER (SpectroSERVER). The number to the right of each bar graph is the total available RAM. CPU Utilization - Displays the percent of total CPU processing power being used by the SpectroSERVER and by other processes running on the machine to which the Performance View is connected. Scroll bars are added to the CPU Utilization graph area to accommodate multiple CPU bar graphs. Disk Utilization - Displays how much of a disk s read/write access capacity is being used. General Data - Provides general data information as follows: SPECTRUM Version - The version of SPECTRUM installed on the server to which the Performance View is connected. SpectroSERVER Uptime - The time in days, hours, and minutes the server to which Performance View is connected has been running. The format for Up Time is: <days>d <hours>h <minutes>m. Number of Clients Connected - The total number of clients connected to the selected SpectroSERVER. Number of Models - The total number of models, including location, organization, and topological models. Number of Accountable Models - All models which are included in the device count calculation (used in SPECTRUM software licensing). See the Search Manager User s Guide (2383) for more information on the predefined search for Accountable Models. Number of Device Models - The total number of physical devices modeled with model types derived from the Device model type. Number of Polled Models - The total number of models that have Polling Status set to true and a non-zero polling interval. Number of Polled Attributes - The current number of attributes being polled by the SpectroSERVER. Performance View Page 77

78 Windows CPU Tabbed Page Number of Logged Attributes - The current number of attributes being logged by the SpectroSERVER. Note: Polled attributes are used to determine if a device is up or down (fault isolation). Logged attributes are used to gather statistical information. Polled attributes may or may not be logged and logged tabulates may or may not be polled. CPU Tabbed Page The CPU tabbed page displays information about the CPU performance and utilization and consists of the CPU Utilization graph, and the Top CPU Users area. CPU Utilization - Always displays the SpectroSERVER usage as a percentage of total CPU capacity. Each system CPU is also displayed. Performance View Page 78

79 Windows Memory Tabbed Page Top CPU Users - Displays the top ten CPU users along with the CPU utilization percentage of each user. Memory Tabbed Page The Memory tabbed page consists of the SS Memory Utilization graph, the Paging Activity graph, and the Top CPU Memory Users area. SpectroSERVER Memory Utilization - Displays how much memory the SpectroSERVER is using. Paging Activity - Displays total paging activity. Top Memory Users - Displays the top ten memory users from highest to lowest along with the memory being used (in MB) by each process. Performance View Page 79

80 Windows Disk Tabbed Page Disk Tabbed Page The Disk tabbed page consists of the Disk Utilization graph, and the Queue Length graph. Disk Utilization - Displays how much of a disk s read/write access capacity is being used. Queue Length - Displays the number of disk access requests pending. Performance View Page 80

81 Windows Network Tabbed Page Network Tabbed Page The Network tabbed page consists of the Network I/O graph, and the Traps graph. Network I/O - VNM read/write bytes only. Does not include traffic from any other sources. Traps Received - Displays the number of unsolicited messages received by the VNM, e.g., SNMP traps. Performance View Page 81

82 Windows Poll/Log Tabbed Page Poll/Log Tabbed Page The Poll/Log tabbed page consists of the Average Poll Latency graph, the Poll Threads in Use graph, the Average Log Latency graph, and the Log Threads in Use graph. Average Poll Latency - Poll latency is the queue time for polling resources to become available. Poll Threads in Use - A poll thread is allocated to every polling operation. Average Log Latency - Log latency is the queue time for logging resources to become available. The Log Latency graph displays the current queue time for the logging process in seconds. Log Threads in Use - A log thread is allocated to every logging operation. Performance View Page 82

83 Windows Health Report Tabbed Page Health Report Tabbed Page The Health Report tabbed page consists of System Capacity pie chart, four action buttons (Start, Stop, Save, Open), a status area, and the main report area. Refer to How To Use Health Report on Page 41 for information about the Health Report tabbed page. Performance View Page 83

84 Windows Timer/Notify Tabbed Page Timer/Notify Tabbed Page The Timer/Notify tabbed page consists of the Average Timer Latency graph, the Timer Threads in Use graph, the Average Notification Latency graph, and the Notification Threads in Use graph. Average Timer Latency - The Timer Latency graph displays the current queue time for the allocation of timer threads in seconds. Timer Threads in Use - This is a rate graph displaying the number of timer threads in use. Average Notification Latency - The Notification Latency graph displays the current queue time for the allocation of notification threads in seconds. Notification Threads in Use - This is a rate graph displaying the number of notification threads in use. Performance View Page 84

85 Windows Preferences Preferences The Preferences window (accessed from the main window s Options menu) lets you customize the following Performance View parameters. Save Settings At Exit - If checked, settings defined during the current session are saved and will be used the next time the application is started. Confirm Exit - If checked, verification is requested by the application before exiting. Show Warnings At Exit - If checked, warnings encountered during this session that have not been resolved will be displayed. Show Server List At Startup - If checked, all servers to which the application can connect are displayed at startup. Otherwise, Performance View connects to the last SpectroSERVER it was connected to. Performance View Page 85

86 Windows Preferences Report When Complete - If checked, the completed report will be automatically ed (in plain text format, not HTML) to the address(es) specified in the Report To field described below. Report To - Enter the addresses to which you want the report sent, separating individual addresses with commas. Title Font - Click the Font button to display the Font Selection window. Label Font - Click the Font button to display the Font Selection window. Chart Line Colors - Click a color button to display the Select Color window. Performance View Page 86

87 Windows Select SpectroSERVER Select SpectroSERVER The Select SpectroSERVER window lets you choose the SpectroSERVER to which you want Performance View to connect. The Filter/Search button lets you find a SpectroSERVER quickly in situations where the Available Servers list is long. Filter - Removes servers from the list that do not match the your criteria. Search - Parses the list and highlights the first server that matches your criteria. Click Next to highlight the next server that matches your criteria. Performance View Page 87

88 Windows Connection Status Connection Status The Connection Status window provides detailed information about the operational status of connection between Performance View and the selected SpectroSERVER such as the original connection time and connection status (connection lost, connection reconnect, change to backup server, return to primary server). Performance View Page 88