Using EMC CLARiiON CX4 Disk-Drive Spin Down with EMC Celerra FAST

Transcription

1 Using EMC CLARiiON CX4 Disk-Drive Spin Down with EMC Celerra FAST Applied Technology Abstract This white paper highlights the ability to use the CLARiiON disk-drive Spin Down feature with EMC Celerra to reduce the power consumption for SATA drives. It provides details of the testing performed to verify whether EMC Celerra FAST is supported by the disk-drive Spin Down feature. Additionally, this paper also covers the recommendations and best practices to implement the feature. April 2010

2 Copyright 2010 EMC Corporation. All rights reserved. EMC believes the information in this publication is accurate as of its publication date. The information is subject to change without notice. THE INFORMATION IN THIS PUBLICATION IS PROVIDED AS IS. EMC CORPORATION MAKES NO REPRESENTATIONS OR WARRANTIES OF ANY KIND WITH RESPECT TO THE INFORMATION IN THIS PUBLICATION, AND SPECIFICALLY DISCLAIMS IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Use, copying, and distribution of any EMC software described in this publication requires an applicable software license. For the most up-to-date listing of EMC product names, see EMC Corporation Trademarks on EMC.com All other trademarks used herein are the property of their respective owners. Part Number h Applied Technology 2

3 Table of Contents Executive summary... 4 Introduction... 4 Audience... 4 Terminology... 4 Technology overview... 5 EMC CLARiiON disk-drive Spin Down... 6 EMC FMA... 8 EMC Celerra FAST... 9 FAST support for CLARiiON disk-drive Spin Down... 9 How EMC tested this technology... 9 Setup details Hardware requirement Software requirement Process overview Assumptions Test details Spin down the CLARiiON Spin down the file system FMA file archival with Spin Down FMA file retrieval with Spin Down FMA storage scaling with Spin Down FMA file retrieval options with Spin Down FMA archived file edit with Spin Down FMA stub file deletion with Spin Down FMA stub scanning with Spin Down FMA stub recovery with Spin Down FMA NAS migration with Spin Down FMA multi-tier archiving over EFD-FC-SATA with Spin Down FMA multi-tier archiving over EFD-SATA-SATA with Spin Down FMA support for storage configurations with Spin Down Best practices No Initial Verify Conclusion References Applied Technology 3

4 Executive summary Data center managers are increasingly demanding the reduction of total cost of ownership (TCO) as corporations ask information technology (IT) and facilities departments to do more with less each year. With increasing electricity and fuel costs, the operational costs of maintaining a data center are on the rise. Sustainable design, global warming, depleting fuel reserves, energy use, and operating costs are gaining in importance in data centers for the following reasons: Large, concentrated use of energy that can be up to 100 times the watts per square foot of an office building Operations running 24/7 have approximately three times the annual operating hours as other commercial properties Backup power generators and uninterruptible power supply (UPS) systems are costly components of a highly available data center. EMC introduced the disk-drive Spin Down feature, an advanced energyefficient technology, with EMC CLARiiON and the EMC Celerra unified storage platform. With this feature enabled, a Spin Down eligible disk drive motor stops running when the drive does not receive any I/O activity for a specific duration. Introduction A study was conducted to determine the feasibility of the CLARiiON disk-drive Spin Down feature as it applies to Celerra unified storage platforms. The objective of this study was to verify whether Celerra file systems that are placed on CLARiiON RAID groups with the Spin Down feature enabled function as expected. This was tested by configuring secondary and tertiary storage locations for the EMC Celerra Fully Automated Storage Tiering (FAST) technology on Celerra file systems, which were created over SATA RAID groups with the CLARiiON disk-drive Spin Down feature enabled. The primary goals of this study were to: Verify that the RAID groups used by Celerra as secondary or tertiary storage with FAST can be quieted for a sufficient duration so that the disk drives in the RAID group are spun down by CLARiiON. Verify that the CLARiiON array spins up disks from low power mode quickly enough to avoid file system timeouts on Celerra when data is sent to the spun-down disks through the FAST technology. This white paper covers a brief introduction of the disk-drive Spin Down feature and its management using Navisphere Manager and Navisphere CLI. It provides detailed test results to show that the Celerra FAST technology supports the disk-drive Spin Down feature of Celerra, along with recommendations and best practices to implement the Spin Down feature. The installation and configuration of EMC File Management Appliance (FMA) are beyond the scope of this document. Audience This white paper is intended for EMC customers, partners, and employees who are considering the use of the CLARiiON disk-drive Spin Down feature. It is assumed that the reader is familiar with CLARiiON storage systems, Celerra unified storage platforms, and their management software. It is also assumed that the reader is comfortable with the usage, configuration, and management of the EMC Celerra FAST technology. Terminology Celerra unified storage platform: EMC network-attached storage (NAS) system. Applied Technology 4

5 Common Internet File System (CIFS): Network file-sharing protocol that is based on Server Messaging Block (SMB) developed by Microsoft. EMC File Management Appliance (FMA): Automates policy-based file archiving. Full power mode: State of a disk drive where the drive has its motor spinning and ready to serve the I/O requests. Fully Automated Storage Tiering (FAST): Technology that moves files from the Celerra file system to archive locations based on predefined policies. Idle time requirement: Specifies how long the Spin Down eligible drives should be idle before the storage system moves them to the standby mode. It is a fixed duration of 30 minutes. Low power mode: State of a disk drive where the motor is no longer spinning. It is also known as the standby mode. Network File System (NFS): Network file-sharing protocol for UNIX-based systems. Spin Down capable RAID group: A RAID group with EMC disk-drive Spin Down feature qualified drives. Spin Down eligible drives: Drives that are qualified for the disk-drive Spin Down feature and satisfy the configuration criteria. Spin Down qualified drives: Drives qualified by EMC to be enabled for the disk-drive Spin Down feature and that can withstand multiple spin ups and spin downs. Technology overview Companies continue to face a deluge of data growth in their environments. The management of unprecedented and unavoidable data growth has been one of the biggest challenges faced by companies today. Based on the manner in which data is organized, the data can be classified as structured (databases), unstructured (file systems), or semi-structured ( ). From earlier studies conducted, it has been observed that most of an organization s data is unstructured (file systems) and a major part of it is inactive on an average. In a growing file storage environment, primary NAS and file servers reach maximum capacity at regular periods of time. Therefore, organizations need to continually purchase more storage. Consequently, organizations may experience other problems as well, such as unplanned server outages due to the running out of space, excessive administrative costs, backup failures, and so on. File tiering helps to address this challenge by moving inactive or infrequently accessed files from the source storage to the archive storage tiers. An intelligent file tiering solution provides a number of benefits such as: Defer additional purchase of storage and lower TCO by moving inactive or infrequently accessed files from a source tier (for example, Fibre Channel) to a target tier of storage (for example, SATA). Reduce backup time when static content is removed from backups. Enable data alignment with retention requirements to meet compliance requirements. Prevent storage administrators from continually dealing with server outages. FMA can be used along with EMC Celerra FileMover APIs for FAST to provide an automated storage tiering solution that moves data from the primary storage to the secondary or tertiary storage based on defined policies. Energy usage also contributes to a significant portion of the facility operating costs of data centers, and hence energy efficiency is an important metric. Data center equipment is supplied power from systems of varying efficiencies and requires cooling infrastructure to support the resulting heat load. Therefore, improving the energy efficiency of data center equipment is a critical step towards achieving an energyefficient data center because the source heat load and the resulting Heating, Ventilation and Air Conditioning (HVAC) system load are reduced. Applied Technology 5

6 Data center equipment has become more energy-efficient through the use of efficient power conditioning equipment and power supplies, lower power processors, and sophisticated power management. Energy efficiency can be defined as the ratio of the energy output to the energy input of a process or a machine. Improved energy efficiency may also improve reliability because it results in less heat being generated within the equipment, and thereby reduces the temperature that components are exposed to. In turn, this improves reliability because the failure rate of electronic components increases with the increase in operating temperatures. Power management helps in improving the energy efficiency of a data center. This is achieved by scaling power consumption so that it is proportional to the workload, that is, the power is reduced when the equipment is idle. This improves the "performance per watt," which is the ratio of the performance of the equipment measured in operations per second to the average input power of the equipment. Most modern servers have built-in technologies that address this requirement. It is recommended to activate power management features to improve energy efficiency. EMC CLARiiON disk-drive Spin Down Celerra unified storage platforms have introduced a new feature called the disk-drive Spin Down feature. When this feature is enabled at the storage system level, the storage system keeps track of how long the drives are idle. If a drive has been idle for more than 30 minutes, the array automatically spins down the drive s motor so that it is no longer spinning. However, the electronics of the drive is still powered up. Thereby, when there is an I/O request for that drive, the array spins the drive up to its normal speed. With this feature enabled, the disk drives can be in any of the following states at a specific instant: Full power The drive is spinning at a normal rate. Low power The drive is in a standby mode, and hence there might be a delay in accessing the data on the drive. Reduced power The drive is transitioning from one mode to another, and hence data cannot be accessed at this point. Support The disk-drive Spin Down feature is available with FLARE release 29. It is not a licensed product and no separate enabler is required to use this feature. Power savings The amount of power that the disk-drive Spin Down feature saves is proportional to how much time a disk spends in standby mode. Internal experiments have shown that low-power SATA II drives save approximately 55 percent power in low power mode as compared to full power mode. The following table shows the power savings in both modes. Table 1 Power savings in low power and full power modes Type of drive Power consumption in Active state (watts) Power consumption in Standby state (watts) Power savings (%) 5400 rpm SATA II rpm SATA II The actual total power savings depends on how long the drives are allowed to operate in a low power state. For further details on power savings provided by low-power SATA II drives and the disk-drive Spin Down feature, refer to the An Introduction to EMC CLARiiON CX4 Disk-Drive Spin Down Technology Applied Technology white paper. Applied Technology 6

7 Criteria for disk drives to enter the standby mode The drives are allowed to enter the standby mode when both storage processors report that the drives have not been accessed for 30 minutes. The array monitors the I/O activity at the RAID group level and not at the LUN level. Therefore, the LUN layout should be designed based on the access profiles of LUNs when using the disk-drive Spin Down feature. With the Spin Down feature enabled on the storage system, the drives with no RAID group created and the drives in a Spin Down enabled RAID group with no LUNs created on them move to a standby mode in less than two minutes. Criteria for disk drives to exit the standby mode When the array receives an I/O request for a drive in standby mode, the I/O operation is kept in a pending state while a request is sent to the drive to exit the standby mode. All further I/O requests to the drive are saved until the drive returns to a ready state. Using drive health check operations, at regular intervals, the array ensures that the drive can be moved from the standby to the ready state whenever the need arises. Drive spin-up algorithm The array allows up to four drives to spin up simultaneously in an enclosure. Also, the array waits 12 seconds before spinning up the next set of four drives in the same enclosure to ensure that the spin-up operations remain within its power budget. Internal tests have confirmed that the latency in moving all drives in an enclosure from standby to ready mode is in the tolerance limits of most operating systems. Background health check services of the array ensure that the usage of the disk-drive Spin Down feature does not compromise data integrity or product reliability. Management The disk-drive Spin Down feature can be enabled or disabled at the storage system level. By default, this feature is disabled on the storage system. This feature can also be enabled and disabled at the RAID group level for further granular management. The disk-drive Spin Down feature settings are persistent across storage system and storage processor reboot operations. The Spin Down feature can be managed from Navisphere Manager and Navisphere CLI commands. For further details on configuration settings and management of the disk-drive Spin Down feature, refer to the An Introduction to EMC CLARiiON CX4 Disk-Drive Spin Down Technology Applied Technology white paper. Monitoring In the Navisphere Manager GUI, special icons are used to represent RAID groups with drives eligible for the disk-drive Spin Down feature and for drives that are in low power mode and in reduced power mode. The array maintains a disk-drive Spin Down log on the storage system when power saving statistics logging is enabled. Power saving statistics logging helps in evaluating the system performance and the impact of the Spin Down feature. Counters are maintained at the drive level such as the instant when statistics were retrieved last time, time in minutes spent by the drive in spinning state (spinning ticks), time in minutes spent by the drive in standby mode (standby ticks), number of times the drive has spun up from standby mode, and the instant at which power saving statistics logging is enabled. Limitations The limitations of the disk-drive Spin Down feature are: Disk-drive Spin Down is supported on CLARiiON CX4 storage systems only, and hence only on Celerra platforms with a CX4 backend. Drives in vault locations are not eligible for Spin Down. Unless all drives in a RAID group are eligible for Spin Down and satisfy the idle time criteria, none of the drives will enter the standby mode. Applied Technology 7

8 A RAID group used for layered applications such as MirrorView, SnapView, MetaLUNs, or Virtual Provisioning is not eligible to use the disk-drive Spin Down feature. In FLARE 29, only 1 TB SATA II drives are qualified for Spin Down. EMC FMA EMC FMA automates policy-based file management by allowing storage administrators to create scheduled policies that move inactive data on a per-file basis to lower tiers of storage such as SATA drives on the same or another Celerra or EMC Centera. This process is done transparently, with no impact to users or applications, and files remain online for fast, easy access when needed. By moving inactive data to a lower-tier storage, FMA helps organizations to reclaim primary storage and lower overall storage costs. FMA can be deployed in two ways: File Management Appliance A standalone device that deploys quickly and easily and provides scalability, high availability, manageability, and flexibility. File Management Appliance/VE A software-only offering that is delivered as a VMware virtual appliance. FMA file operations When an archive task is executed, FMA scans source file systems through CIFS or NFS and searches for files that meet a predefined policy. When a file matches a policy, it is written to the specified archive tier of storage. After the write is committed, the file at the source location is replaced with a small stub file of 8 KB in size that contains all the data necessary for file recall. When a client requests an archived file, FMA fetches the location of the archived file from its stub file, which contains the required metadata. FMA requests the file from the archive storage and retrieves it. If the source tier is Celerra and the target tier is a CIFS or NFS target, Celerra requests the file and directly reads it through the CIFS or NFS protocol. The multi-tiered archiving feature of EMC FMA enables the creation of policies that move the archived files off the secondary tier to a tertiary tier. With this operation, the stub files remain on the primary storage and are updated with the new file location from the tertiary storage. Management FMA offers multiple reporting and logging capabilities, such as: Viewing archiving history by capacity, average file size, or file type Grouping file servers for reporting Filtering reports by time or file server Generating full-color pie charts that provide a snapshot of files archived by size and type Support FMA is integrated with Celerra and EMC Centera APIs so that the existing functionality and processes can be maintained. Compliance features available in the underlying storage can be leveraged. When creating an archiving policy, the administrator can specify a retention period that will be set when archiving a file to a Celerra File-Level Retention enabled file system, EMC Centera Compliance Edition Plus, or other supported storage targets. Availability FMA is designed for enterprises with enterprise-level availability and scalability. It operates out of the data path and does not rely on a metadata database, so performance is not affected and data integrity does not become an issue. A disaster recovery strategy specifically to protect metadata is not required because all the information needed to recall a file is maintained in a small stub file. Applied Technology 8

9 EMC Celerra FAST EMC Celerra Fully Automated Storage Tiering (FAST) provides an easy and cost-effective solution for IT managers to provide the right level of service to critical applications. By simply pooling storage resources, defining a policy, and applying it to an application, FAST automatically and dynamically moves data across Storage Types, so that it is in the right place at the right time. FAST, based on FMA and Celerra APIs, transparently relocates files to the appropriate Storage Type. FAST automatically moves files from one Storage Type to another, that is, from primary high-performance Enterprise Flash Drives (EFDs) to secondary Fibre Channel (FC) drives and then to tertiary low-cost Serial ATA drives (SATA) based on the activity level of the file by using FMA. FAST offers tiering in-the-box and outside-the-box: With in-the-box tiering, two or three storage tiers can be defined in a single Celerra storage system. These tiers can be Flash, FC, and SATA storage with varying levels of RAID protection. With outside-the-box tiering, EMC Centera and Atmos based storage can be used for a more costeffective tiering of files. These options and configurations are transparent to users and applications as they continue to access the file with the same path name from the source location. FAST support for CLARiiON disk-drive Spin Down Celerra file systems created on low-power SATA II drives were configured as secondary and tertiary locations for the EMC Celerra FAST technology. From the tests, it was observed that it was possible to quiet the SATA II disk drives, over which the Celerra file systems were created, long enough so that the disk drives could be spun down and moved to low power mode. It was also observed that the SATA II drives placed in low power mode can be spun up quickly enough during archive file recall operations to avoid any file system access timeout situations. Hence, the FAST technology supports the CLARiiON diskdrive Spin Down feature on Celerra. The following features of the EMC Celerra FAST technology were supported when Celerra file systems placed on Spin Down enabled SATA II disk drives were used as secondary and tertiary storage locations: FMA file archival and retrieval operations to Spin Down enabled Celerra file systems. FMA file edit operation on files archived to Spin Down enabled Celerra file systems. FMA file retrieval operations with the DHSM parameter read_policy_override set to full/none on the source file system. FMA stub file deletion with archived files placed on spun-down SATA drives. FMA stub scanning operation with archived files placed on spun-down SATA drives. FMA stub recovery operation with archived files placed on spun-down SATA drives. FMA NAS migration operation with archived files placed on a set of spun-down SATA drives to another set of spun-down low-power SATA II drives. FMA multi-tier archiving operation with the secondary storage location on FC or SATA II drives and the tertiary location defined on a set of SATA II drives in low power mode. FMA support for different RAID configurations with the Spin Down feature enabled on RAID groups. How EMC tested this technology An EMC Celerra NS-480 was used for primary, secondary, and tertiary storage in the FAST environment. For FAST secondary and tertiary storage, Manual Volume Management (MVM) was used to configure Celerra file systems from CLARiiON 4+1 RAID 5 SATA II groups that had the Power Savings feature enabled. Standard best practices were followed for Celerra Data Mover and network configurations. Several tests have been performed to verify the feasibility of the disk-drive Spin Down feature of FLARE release (R29). The following sections provide the details of system configuration, tests performed, and observations from resulting behavior. Applied Technology 9

10 The objectives of this testing are: To characterize the effect of the increasing number of low-power SATA II RAID groups on file archival and retrieval operations using FAST. To verify if there is any adverse effect of moving the SATA II drives to low power mode on various features of EMC FMA. Setup details This section provides a description of the test environment along with the hardware and software used to perform the testing discussed in this paper. The details of the test environment are: The test bed consisted of an FMA and a CIFS client configured on Dell 1950 hardware, which was used to access the FMA user interface and to perform file archival and retrieval operations. A Celerra NS-480 was used to provision appropriate tiers of storage for FAST. For storage configuration, 4+1 RAID 5 groups were created on low-power SATA II drives and LUNs were bound. The respective disk volumes were striped across to create a stripe volume and a metavolume was configured over which a file system was created. This file system was configured as a secondary or tertiary storage location for archiving files from the source or primary location. The FAST environment was set up using FMA and Celerra File Mover APIs. An EMC FMA version build was used, which has the multi-tiering functionality to move data from one Storage Type to another. A CIFS client was used to test the FAST technology by using three types of storage: EFDs as primary storage, FC drives as secondary storage, and low-power SATA II drives as secondary and tertiary storage in the same EMC Celerra storage system. Hardware requirement The following table lists the hardware components used to perform this testing. Table 2 Hardware configuration Component Description Usage Celerra NS40G Gateway system with 2 Data Movers NAS storage for FAST (one primary and one standby) primary, secondary, and CLARiiON CX4-480 EMC FMA 5 EFDs 5 FC drives 30 SATA II drives tertiary locations Back-end storage FMA Dell PowerEdge 1850 Intel Xeon CPU, 3.00 GHz, 4 GB RAM Windows CIFS Client Applied Technology 10

11 Software requirement The following table lists the software components used to perform this testing. Table 3 Software configuration Component Description Usage Celerra DART CMR 10 CLARiiON FLARE Release 29 (Minimum version that includes support for disk-drive Spin Down) EMC FMA FMA Operating system Windows Server 2003 R2 Enterprise x64 Edition Service Pack 2 Windows CIFS Client Process overview Tests were performed to verify the feasibility of the CLARiiON disk-drive Spin Down feature with EMC Celerra. The FAST technology was tested with one to six RAID groups of Spin Down eligible SATA II drives. For all tests executed, the following statistics were recorded for analysis and documentation purposes: Archival duration with disks in full power mode Archival duration with disks in low power mode Retrieval duration with disks in full power mode Retrieval duration with disks in low power mode Tests were conducted with disk drives in low power and full power modes to determine the amount of additional delay in archival and retrieval operations. In addition, tests were performed to verify the impact on FMA features when SATA II drives are moved to low power mode. Assumptions The assumptions for the tests were: Before the file archival or retrieval operation is initiated, all selected test drives are in low or full power mode as per the test requirement. During the archival or retrieval operation, all selected test drives are spun up to full power mode from low power mode. Test details This section provides details of the tests performed to evaluate the functionality and performance of FAST when SATA II drives were used in low power mode as a storage location for archived files. All the test cases used the DHSM parameter setting read_policy_override as none on the source file system, which does not override the migration method in the stub file. The default read policy for the stub files was read_pass_through. Therefore, read recalls of files did not copy the files contents back to the primary or production file system during the tests. For all the tests, the Power Savings option was enabled on the CLARiiON storage system and all RAID groups were created on low-power SATA II drives. The disk power savings statistics logging option was also enabled. Applied Technology 11

12 Unless specified otherwise, all archival and retrieval tests were performed with 100 files, each of the size 10 MB, populated with 100 percent random data. The following file systems created on EMC Celerra were used as storage locations: EFD_FS A file system created on a 4+1 RAID 5 group over EFDs. This was used as the source or primary location for all tests. FC_FS A file system created on a 4+1 RAID 5 group over FC drives. This was used as the secondary location for some of the tests. SATA_FS A file system created on a specific number of 4+1 RAID 5 groups over low-power SATA II drives. This was used as either a secondary or tertiary storage location in the tests. For all the tests, the storage locations were accessed over the CIFS protocol from FMA. On the primary file system, that is, the file system on EFDs, Distributed Hierarchical Storage Management (DHSM) was enabled to act as a source file system. The file systems created on FC and SATA drives were configured as repositories in FMA to place the files archived from the source location. The DHSM connection was configured on Celerra to all storage types for CIFS. For all the tests, the source files to be archived were placed on a file system EFD_FS that was created over five EFDs. One hundred files, each of the size 10 MB, were used towards archival/retrieval operations. Spin down the CLARiiON The objective of this test was to validate that Celerra allows a CLARiiON RAID group created over lowpower SATA II drives to achieve low power mode when no LUN is bound on the RAID group. A 4+1 RAID 5 group was created on Spin Down eligible SATA II drives. No LUNs were bound on the RAID group. Subsequently, the Power Savings option was enabled on the RAID group. The RAID group moved to a standby mode in approximately less than two minutes after the Power Savings option was enabled. Spin down the file system The objective of this test was to validate that Celerra allows a CLARiiON RAID group created over lowpower SATA II drives to achieve low power mode when the Celerra file system placed on the RAID group is not accessed by any client for an extended period of time. A 4+1 RAID 5 group was created on Spin Down eligible SATA II drives. A CLARiiON LUN was created on the RAID group. Using the MVM configuration, a 100 GB file system was created. After the file system was created, the Power Savings option was enabled on the RAID group. After 30 minutes, the RAID group moved to low power mode. Because there were no users accessing the file system and thereby the drives in the RAID group, it was observed that the RAID group moved to low power mode within 30 minutes after the file system creation. FMA file archival with Spin Down The objective of this test was to validate if the FAST technology supports the archival write operation to low-power SATA II drives in the spun-down state. Applied Technology 12

13 A SATA_FS file system was created on five Spin Down eligible SATA II drives. A CIFS share was created over the file system. SATA_FS was configured as the secondary destination location for archival. The file system was accessed through the CIFS protocol. A policy was created to archive all files from the source location with a file size greater than 8 KB. A schedule was executed to run the policy with the run now option. It was confirmed that the files were archived and moved to a defined secondary location. The source files were replaced with stub files each of the size 8 KB. The following behavior was observed: SATA drives were spun up to full power mode. All source files that satisfied the policy criteria were archived and moved to the SATA_FS file system. It was confirmed that FMA can tolerate the latency while spinning up the five SATA drives to full power mode during the write operation. The file archival test was successful. FMA file retrieval with Spin Down The objective of this test was to validate if FAST supports file retrieval from the archive location while the low-power SATA II drives are in a spun-down state. One hundred source files on EFD_FS were archived to SATA II drives and replaced with stub files. The I/O activity on SATA drives was stopped, and the drives were allowed to move to low power mode. The drives were allowed to remain in low power mode for 10 minutes. A copy operation from EFD_FS to a local directory on the CIFS client was initiated. This operation initiated an FMA file retrieval operation from the secondary location. Based on DHSM settings on the Data Mover, the stub files at the primary location will not be replaced by the data files as a result of the read operation. The following behavior was observed: SATA drives were spun up to full power mode. Required data blocks were read from the archived location on SATA drives and copied to the specified local directory on the CIFS client. It was confirmed that FMA can tolerate the latency while spinning up the five SATA drives to full power mode during the read operation. The archived file read test was successful. FMA storage scaling with Spin Down The objective of this test was to check the effect of enabling the Power Savings option on an increasing number of SATA II RAID groups over FMA file archival/retrieval operations. Thirty Spin Down eligible SATA II drives were configured with 4+1 RAID 5 groups. The Power Savings option was enabled on all RAID groups. A single LUN of 100 GB in size was created from each RAID group. LUNs were made accessible to Celerra and marked as disk volumes. A single slice volume of 20 GB was created from each disk volume. Based on the test requirement, multiple slice volumes were striped across with a 64 KB stripe element size to create a stripe volume over which a metavolume was created. A file system was created from the metavolume. A CIFS share was configured over the file system. An FMA archival location was configured on the CIFS share. The FMA policy was scheduled to archive 100 files of 10 MB size each to the CIFS share. Applied Technology 13

14 With SATA drives in full power mode, the file archival test was initiated and the time taken to perform the operation was recorded. All the caches in the I/O path from the CIFS client to the storage were cleared. The I/O activity on SATA drives was stopped, and the drives were allowed to move to low power mode. The drives were allowed to remain in low power mode for 10 minutes. The file archival schedule as defined in FMA was initiated. This operation spun up the SATA drives to perform the writes as a result of the file archival operation. It was observed that the files were archived to the CIFS share and the time taken to perform the operation with SATA drives in low power mode was recorded. All caches in the I/O path from the CIFS client to the storage were cleared. With SATA drives in full power mode, a copy of the archived files was done from the CIFS share to the local directory on the CIFS client. This action initiated a read operation from SATA drives. The time taken to complete the copy operation with SATA drives in full power mode was recorded. All caches in the I/O path from the CIFS client to the storage were cleared. The I/O activity on SATA drives was stopped, and the drives were allowed to move to low power mode. The SATA drives were allowed to remain in low power mode for 10 minutes. A copy operation of archived files was done from the CIFS share to the local directory on the CIFS client. This action spun up the SATA drives to full power mode to serve the read request. The time taken to complete the file copy operation with SATA drives in low power mode was recorded. FMA file archival and retrieval operations were successfully completed with six RAID groups, that is, 30 SATA II drives placed in low power mode. FMA sustained the latency caused by spinning up the 30 drives before performing the read and write operations. The following figure shows the time taken by FMA file archival operations with drives in low and full power modes and with varying number of 4+1 RAID 5 groups. Figure 1 FMA file archival latency with varying number of RAID groups As shown in Figure 1, an approximately similar latency overhead was observed with the increasing number of RAID groups placed in low power mode. This behavior can be explained from the fact that the Celerra file system was mounted with the default options (cached and prefetch), and the write I/Os may have been cached while drives in the low power mode were spinning up their motors. The following figure shows the time taken by FMA file retrieval operations with drives in low and full power modes and with varying number of 4+1 RAID 5 groups. Applied Technology 14

15 Figure 2 FMA file retrieval latency with varying number of RAID groups As shown in Figure 2, a continually increasing latency overhead was observed with the increasing number of RAID groups placed in low power mode. The reason for this behavior can be explained as follows. Irrespective of the number of drives in low power mode, the array spins up only four drives at a time from low power to full power mode. The array waits for 12 seconds before spinning up the next set of four drives. Hence, it takes longer to spin up more numbers of drives from low power mode. In the current environment, because the SATA drives over which the file system was placed was inactive for at least 30 minutes before the backup/restore operation, the I/O request could not be served from the Data Mover cache, thereby forcing all drives to spin up before the FMA file retrieval operation could be performed. FMA file retrieval options with Spin Down The objective of this test was to validate whether the spun-down drives have any effect on the FMA functionality with non-default read_policy_override options. With FMA, the source files when archived will be replaced by stub files, and the actual data will be moved to the secondary location. During a read operation on archived files, the data can be accessed from the primary or secondary location based on the read_policy_override parameter setting: If the parameter is set to none, data will be read directly from the secondary location to serve the read request without any changes to the stub files at the source location. This is the default setting. If the parameter is set to full, the stub files at the source location will be replaced by the actual data files from the secondary or tertiary location before serving the read request. The procedure for this test was as follows. The data set from the source location was archived to a secondary location using FMA. The DHSM setting for the read_policy_override parameter was modified to full on the primary file system. The I/O activity on the SATA file system was stopped, and the drives were allowed to move to low power mode. A copy operation of archived files from the source location to a local directory on the CIFS client was initiated. This operation moved the data files from the secondary to the primary location and replaced the stub files. SATA drives were spun up during the process. This test was performed with two data sets a single file of 1 GB in size and 100 files, each of 10 MB in size. Applied Technology 15

16 With SATA II drives in low power mode, it was observed that there was no effect on the time taken to perform the read operation when the FMA read_policy_override was set to full or none on the primary file system. FMA archived file edit with Spin Down The objective of this test was to verify whether the spun-down drives had any effect on the edit operation performed on the archived files using FMA. With FMA, when write operations are performed on archived files, the stub files at the source location will be replaced by the actual data files copied from the secondary location. Because the behavior is similar to the FMA environment with the read_policy_override set to full, it can be assumed that the SATA drives in low power mode have no effect on the archived files edit operations. FMA stub file deletion with Spin Down The objective of this test was to check whether the spun-down drives had any effect on the deletion of stub files created during file archival using FMA. With FMA, upon archival, the source files will be replaced by stub files while the actual data is moved to the secondary location. FMA policies can be defined to delete the stub files on satisfying specific criteria such as files not accessed for a specific period. The data set from the source location was archived to a secondary location using FMA. The I/O activity on the SATA file system was stopped, and the drives were allowed to move to low power mode. A policy was scheduled to delete the stub files. This action permanently deleted the stub files from the source location along with the archived data files at the secondary location. SATA drives were spun up during the process. The SATA drives in low power mode had no effect on the FMA stub file deletion operation. FMA stub scanning with Spin Down The objective of this test was to validate whether the spun-down drives had any effect on scanning the stub files created during file archival using FMA. With FMA, the stub files at the source location can be scanned to verify that all data files at the secondary or tertiary location have the respective stub files available at the source location. The archived files with no stub files available at the source location are marked in the database during the stubs scanning process. The data files that do not have stub files available for more than 30 days at the secondary location will be marked as orphan files. FMA policies can be defined to delete the orphan files if required. The data set from the source location was archived to a secondary location using FMA. The I/O activity on the SATA file system was stopped, and the drives were allowed to move to low power mode. The stub files were manually deleted from the source location. This operation had no effect on the archived files available at the secondary location. Hence, the SATA drives did not spin up. A policy was scheduled to scan the stub files at the FMA source location. This action marked the secondary files that had the respective stub files missing at the primary location. Applied Technology 16

17 The SATA drives did not spin up to full power mode during this process and thereby had no effect on the stub scanning operation. As expected, the stub files that were deleted at the source location did not get regenerated as a result of this operation. FMA stub recovery with Spin Down The objective of this test was to validate whether the spun-down drives are forced to move to full power mode while performing the recover stub operation on archived files using FMA. FMA offers protection against the accidental deletion of stub files, which points to archived files available at the secondary location. In this scenario, FMA provides an option to re-create the stub files for the respective archived files using the recover stub feature. The data set from the source location was archived to a secondary location using FMA. The I/O activity on the SATA file system was stopped, and the drives were allowed to move to low power mode. The stub files were manually deleted from the source location. This operation had no effect on the archived files available at the secondary location. Hence, the SATA drives did not spin up. A policy was scheduled to recover the stub files at the FMA source location for a selected set of files. The stub files for a selected set of archived files were re-created at the source location. SATA drives were not spun up to full power mode as a result of this operation. FMA NAS migration with Spin Down The objective of this test was to verify whether the spun-down drives had any effect on the NAS migration feature of FMA. With FMA, the archived data files can be migrated from one secondary repository to another without affecting the stub files at the primary location or the user access to the data files. This feature of FMA allows the migration of data from one secondary location to another and automatically updates the stub files with the new location in the primary file system. The data set from the source location was archived to a NAS repository created on a set of low-power SATA II drives using FMA. A second NAS repository was configured on another set of low-power SATA II drives. The I/O activity on the SATA file systems was stopped, and the drives were allowed to move to low power mode. Using the FMA NAS migration feature, a policy was scheduled to migrate archived data files from the first NAS repository to the second. As a result of this operation, the data files were physically moved across the NAS repositories and the stub files at the source location were updated accordingly. The SATA drives over which both the NAS repositories were created were spun up during the process. The FMA NAS migration feature can tolerate the latency overhead caused from spinning up the SATA drives over which the source and target NAS repositories were placed. FMA multi-tier archiving over EFD-FC-SATA with Spin Down The objective of this test was to validate the effect of SATA drives in low power mode when used as tertiary storage with FMA. With FMA, the archived data files available on secondary storage can be further moved to tertiary storage based on policies defined in FMA. During this process, the archived files were physically moved from secondary to tertiary storage and the stub files at the source location were updated. All further user access Applied Technology 17

18 to archived files will be redirected from the source file system to the tertiary file system and all the storage space held by archived files at the secondary location will be released. The user who accesses the files from the primary file system continues to access them regardless of which repository the archived files are moved to. The data set from the source location was archived to a NAS repository created on a set of FC drives as a secondary location using FMA. A second NAS repository was configured on a set of low-power SATA II drives. The I/O activity on the SATA file systems was stopped, and the drives were allowed to move to low power mode. Using the FMA multi-tier archiving feature, a policy was scheduled to migrate the archived data files from the secondary location to the tertiary location, which is the NAS repository based on SATA drives. As a result of this operation, the data files were physically moved across the NAS repositories, and stub files at the source location were updated. The SATA drives were spun up during the process. FMA multi-tier archiving can be successfully configured with FC drives used as secondary storage and SATA drives used as tertiary storage. FMA multi-tier archiving over EFD-SATA-SATA with Spin Down The objective of this test was to verify the effect of SATA drives in low power mode when used as both secondary and tertiary storage with FMA. The data set from the source location was archived to a NAS repository created on a set of low-power SATA II drives as the secondary location using FMA. A second NAS repository was created on another set of low-power SATA II drives. The I/O activity on the SATA file systems was stopped, and the drives were allowed to move to low power mode. Using the FMA multi-tier archiving feature, a policy was scheduled to migrate archived data files from the secondary location to the tertiary location. As a result of this operation, the data files were physically moved across the NAS repositories and the stub files at the source location were updated. The SATA drives were spun up during the process. FMA multi-tier archiving can be successfully configured with low-power SATA II drives used as both secondary and tertiary storage. It was observed that EMC FMA can tolerate the latency overhead from spinning up the SATA drives used as secondary and tertiary storage. FMA support for storage configurations with Spin Down The objective of this test was to validate whether FAST supports different RAID configurations such as 12+2 RAID 6 on low-power SATA II drives. A SATA_FS file system was created on 14 Spin Down eligible SATA II drives configured as a 12+2 RAID 6 group. A CIFS share was created over the file system. SATA_FS was configured as the secondary destination location for archival. The file system was accessed through the CIFS protocol. The I/O activity on the SATA drives was stopped, and the drives were allowed to move to low power mode. A policy was scheduled to archive all files from the source location with a file size greater than 8 KB. The files were archived and moved to the defined secondary location with the SATA drives in low power mode. The I/O activity on the SATA drives was stopped, and the drives were allowed to move to low power mode. A copy operation of data files was initiated from the source location to the local directory on the CIFS client. SATA drives were spun up to full power mode during both archival and retrieval operations. It was confirmed that FMA can tolerate the latency due to the spinning up of the 14 SATA drives to full power mode during the read and write operations. Applied Technology 18

19 Best practices The array monitors the I/O activity at the RAID group level and not at the LUN level. Therefore, the LUN layout should be designed based on the access profiles of LUNs when using the Spin Down feature. No Initial Verify LUN Initial Verify is an array background service that helps to detect and correct potential disk media errors before they become unrecoverable. This feature can be configured on individual LUNs during creation, that is, the LUN binding process. This option, when set, initiates a background process to check all parity sectors in the LUN for any inconsistencies. It was observed that when LUNs are created without the No Initial Verify option selected, the Initial Verify background service caused a low bandwidth read activity on the disk drives over which the LUN was created. It takes a couple of days to complete this initial verify operation even on a reasonably sized RAID group, for example, a 4+1 RAID 5 group. Due to this ongoing background I/O on the drives, the RAID groups created on them were not moved to low power mode until the Initial Verify operation was completed, even though there were no I/O requests from user applications. After the Initial Verify operation is completed, this option has no effect on the LUNs and the switching process of RAID groups between full power mode and low power mode. If No Initial Verify option is selected during the LUN creation, after 30 minutes of I/O inactivity on the drives, the RAID groups enabled for the disk-drive Spin Down feature move to the low power mode immediately. Hence, it is recommended that the No Initial Verify option should not be selected during the LUN creation, keeping in mind the long-term perspective of data accessibility and correctness. Conclusion This white paper provides a brief overview of the EMC CLARiiON disk-drive Spin Down feature and its support on EMC Celerra unified storage platforms. Testing was performed to verify that the feature is supported by FAST using EMC FMA, and that file archival and retrieval operations can be successfully performed during automated storage tiering on EMC Celerra Spin Down enabled drives both in full power and low power modes. References The following documents provide additional relevant information about FAST and CLARiiON disk-drive Spin Down. Access to these documents is based on your login credentials. If you do not have access to the content listed below, contact your EMC representative: An Introduction to EMC CLARiiON CX4 Disk-Drive Spin Down Technology Applied Technology Using EMC CLARiiON CX4 Disk-Drive Spin Down with EMC Celerra Applied Technology EMC File Management Appliance and File Management Appliance/VE Version Getting Started Guide Using EMC Celerra FileMover Technical Module EMC Celerra FAST Reference Architecture Automated Storage Tiering on EMC Celerra Reference Architecture EMC Celerra FAST Validation Test Report Automated Storage Tiering on EMC Celerra Validation Test Report Applied Technology 19