Continuous Processing versus Oracle RAC: An Analyst s Review EXECUTIVE SUMMARY By Dan Kusnetzky, Distinguished Analyst Most organizations have become so totally reliant on information technology solutions that it simply can t function without them. This means that these organizations are directed to consider ways to keep applications running continuously. Business needs and constantly changing regulations, however, are forcing these organizations to change how their information technology solutions are being used and their expectations about the scalability and scope of those solutions. Organizations have to consider the impact of these changes from three different perspectives on purchase criteria. This includes an economic, a technical as well as a functional vantage point. For decades organizations have turned to traditional high availability solutions that are based upon clusters of systems harnessed together to address the needs of a single workload. While this approach has been workable, it has brought with it the requirement for and cost of multiple systems, high speed networking, storage servers, clustering software, clustered database engines and applications architected to use clustering as part of their normal operations. Multi-tier, distributed applications must be in totally synchronization or data can be lost. This adds an additional challenge to creating highly available application environments. Managing nodes and the workloads they are supporting is one level of complexity. The need to manage state between layers of application software that are spread over multiple systems brings the complexity to a much higher level. Another approach has been available for decades as well. This approach centers on the deployment of purpose-built fault tolerant/continuous processing servers rather than focusing on software-based solutions. Since this approach does not rely on or need clustering technology to work, it might be a better, more cost effective and simpler solution to the requirement that applications and workloads always be available for use. This paper focuses on an analyst s review of different approaches to making a database-based workload continuously available. THE PROBLEM Although organizations would prefer to focus on their core business, changes are constantly being forced on them. Regulations require that more data be kept and it be kept for a longer period of time. Applications are also becoming increasingly complex, multi-layer, multi-system collection of servers that might be spread all over an organization s data center or, perhaps to many data centers. Organizations find themselves facing a data center that contains both established and new system types that include mainframes, midrange systems, industry standard X86 systems, Organizations are discovering that their applications need faster access to larger amounts of data. These organizations are also discovering that more people need access to this data, including staff members, contractors, and customers. Document #20120805 The Kusnetzky Group 2012 The Kusnetzky Group is an independent supplier of marketing services to suppliers end user organizations and suppliers in the systems, virtualization and open source technology markets. The opinions presented in this document are based upon our research, our personal experiences and actual use of technology regardless of whether this document or the supporting research were sponsored by one or more of the Kusnetzky Group s clients. Stratus Technologies is distributing this report with Kusnetzky Group s permission. This document may not be copied in whole or in part without the written permission of the Kusnetzky Group.
The need to access applications and data has become so extreme that there simply is no time for a failure or an application slow down. THE ECONOMIC VIEW If we consider the economic view, when an application, workload or, perhaps, entire system slows down or becomes unavailable, the organization can experience a number of costs. The first of which is the potential for lost business. When different approaches to create continuously available information technology solutions are considered, other costs come to the forefront. These costs include the expense of acquiring and deploying multiple systems as well as the staff-related costs of operations, administration, technical support, training and the like. Other important considerations include the costs of redundant systems, their memory, storage, networking components and software licenses for each system. A more subtle, but equally important, area of costs when systems do not operate properly are those related to the following: Rework of damaged goods Disposal of waste created by processes that are managed by failed information systems Legal liability for staff or customer injuries due to the failure The need for expedited shipping for finished goods The impact of lost sales Losing customers due to time to market failures THE TECHNICAL VIEW From a technical vantage point, the considerations revolve around making the best selection form a number of different approaches to providing continuously available computing solutions. The key questions to focus on here are the availability of the proper systems, software and staff expertise available? Since the technical buyer is often also focused on compliance, the technical specifications must comply with: Industry standards Requirements for performance Requirements of service level agreements Interoperability standards Compliance with regulations THE FUNCTIONAL VIEW The functional perspective centers upon the business requirement that things need to simply work. The organization can t afford unexpected downtown or applications that run so slowly that they re of little use. The functional buyer is concerned about getting the best tool for the job. This decision maker also considers organization limitations such as lack of specific skills, available time and both political and sociological issues of the organization. It is possible that technical solutions are available that are not workable within the context of the organization. THE OVERALL VIEW Small to medium size organizations may face a much more limited staff and set of skills this can mean that a single individual must consider all of these things. #20120805 The Kusnetzky Group 2012 Page 2
DIFFERENT APPROACHES TO A SOLUTION Organizations have been turning to traditional multi-system approaches to address these issues. They ve harnessed together the power of many individual systems using clustering technology. Each system must have its own operating system, database engine, applications, management software and networking tools. Organizations that have selected clusters as their basic approach to providing continuous processing may have done so at the suggestion of the operating system or database supplier. They might not have known that there is another time-tested approach available the deployment of fault tolerant/continuous processing systems. THE PAIN THESE PROBLEMS CAUSE The pain these problems cause really depend upon the organization, its core capabilities and the approach it has taken to deal with the problems. Let s examine this from three different vantage points. THE ECONOMIC VIEW If we adopt the economic view of the clustering approach, we need to acknowledge that the organization will be required to purchase a number of systems, the operating systems required to make these systems functional, the clustering software that harnesses them together, the database software that supports the applications and the applications for each member of the cluster. Furthermore, other cost factors should be considered as well. Storage servers and high-speed networking equipment must also be added to the equation. All in all, this can be quite expenses in terms the necessary hardware, software and the different types of staff expertise required to make this approach functional properly. A key question to consider when thinking about using clusters as an approach to application availability and reliability is does clustering represent the most economical solution? It all depends. THE TECHNICAL VIEW From a technical vantage point, clustering is a complex solution at best. The organization must have the staff expertise in systems, operating systems, database engines, applications, and clustering technology. It would be wise to consider whether the staff has the expertise to install and operate a complex clustered configuration. If another simpler approach existed, would it, in all likelihood, would be a better choice. THE FUNCTIONAL VIEW Clustering is a complex solution and it would be wise to consider if this approach makes it easier for the organization to deal with today s business requirements for continuously available applications. If it were possible to use a single system to replace the multi-system solution, it would make far more sense. FINANCIAL SERVICES VIEWPOINT Transactions in a financial services environment can be worth a great deal of money, losing even one could be very costly. Financial services is a highly regulated market as well. Organizations must keep more data, keep if for a longer period of time and allow more interested parties access to this data. So a slow down or outage can be very costly. One point to consider in a financial services environment is how quickly a new technology solution can pay for itself. Since organizations in this market are trying to reduce overall costs while providing necessary solutions, it is wise for decision makers to always be considering better solutions. #20120805 The Kusnetzky Group 2012 Page 3
HEALTHCARE VIEWPOINT Patient and operational data must be protected. Losing a transaction or losing data could literally lead to loss of life. As with financial services, this market is highly regulated and some of the same concerns apply. More data must be kept for longer periods of time and more interested parties must be allowed to examine that data. Unlike in the financial services market, healthcare suppliers must live within very tight cost control policies. The adoption of new technology happens far less frequently and must pay for itself over a longer period of time. Finding a solution to making applications and data highly available and reliable that has a rapid return, is clearly worth being examined. TRADITIONAL APPROACH The traditional approach is to harness together multiple systems to provide a single computing solution. These configurations are called clusters Systems are connected together using some form of high speed interconnect (see Figure 1.) Optionally, the clustered systems may share network attached or local storage. Some approaches require shared network attached storage. Figure 1: Clustered Systems There are many different types of clusters. The physical configuration often looks identical. The primary difference is the type of software being used to create the clustered environment. There are two different forms of processing virtualization clusters. One depends upon a cluster manager that makes it possible to see multiple systems as a single computing resource (see Figure 2.) The cluster manager may be part of the operating system or part of a database engine depending upon the application s requirements. If we consider Oracle database-based applications, the cluster manager is an optional database product called Oracle RAC (reliable application clusters) If the cluster manager detects a slow down or failure, applications may be moved from one system to another. If the database is clustered, database transactions may be moved from one system to another. #20120805 The Kusnetzky Group 2012 Page 4
Figure 2 Processing Virtualization Cluster Manager Clustered solutions require multiple systems, storage virtualization and, quite often, applications written to be cluster aware. Implementing this form of processing virtualization requires expertise in clustering. If the organization is using an application that was written to be cluster-aware, data will not be lost if there is a failure. It can take quite some time, however, for this form of cluster to respond to an outage. The cluster manager has to examine each process that is executing, determine what cluster resources it is using and then assign that application to one of the remaining systems. Furthermore, multi-tier, distributed application components must be aware of what the cluster manager is doing so that the state of each component can be maintained when the cluster addresses a failure scenario in progress. Otherwise data can be lost. While this approach creates scalable, reliable solutions, the response to a failure can be time consuming. This approach is complex and imposes higher hardware and software costs as well as the requirement for a great deal of staff expertise. It might be wise to consider another approach. WHAT WOULD THE IDEAL APPROACH BE? An ideal approach would be simple. It would be based upon a single system, a single storage system, only require a single set of licenses (operating system, database, application) and not require the complexity and expense of clustering. Furthermore, application frameworks and application components must not be required to be aware of cluster manager functions. The ideal approach would also be easy to install and operate and not require so many different types of expertise be on staff. Everyone would be prevented from seeing or experiencing the result of hardware failures. When that was not possible, the few failures that appeared would be addressed nearly instantaneously and would not require staff intervention to: Recover state of network Recover state of storage services Restart application frameworks Restart applications #20120805 The Kusnetzky Group 2012 Page 5
Restart database engines and execute processes to bring the database back up to date Printing and batch queues restarted and allowed to processes applications through the point of failure and back to normal execution. Because the ideal solution is simple, it would have a lower cost of ownership as well as offering a much more rapid return on investment. FAULT TOLERANT/CONTINUOUS PROCESSING SYSTEMS The industry developed an approach that represents this idea years ago. It is known in the industry as a fault tolerant or continuous processing system. Special hardware has been designed to prevent failures from being seen. These systems include multiple processors, memory, networking and storage hardware. These systems address hardware failures automatically at system speed. These systems prevent disasters rather than merely recovering from them. Solutions based upon these systems are significantly simpler than clustered solutions. They are based upon a single system that executes a single operating system, database engine and application If we consider a simple cost comparison model, this approach is less costly to utilize. It also has the benefit of taking less data center space, using less cooling and less power. TABLE 1 SIMPLE COST COMPARISON MODEL Clustered Solution Fault tolerant/continuous processing solution Systems Multiple Single Processors Multiple Multiple (in a single enclosure) Operating system Multiple Single Database engine One for each server Single Clustering Oracle RAC Not needed software Storage Requires use of storage server Storage server not required Space Power and cooling Expertise Space for multiple systems, storage server and networking equipment Power and cooling for multiple systems, storage server and networking equipment - Operating system - Applications - Database engine - Clustering software - Clustered system management Space for single system, its storage and networking Power and cooling for a single system, its storage and networking - Operating system - Applications - Database engine - Single system management CUSTOMER COMMENTS Allegheny County s Emergency Services 9-1-1 Division handles 4,500 to 5,000 calls for police, fire and EMS dispatch services each day. Preventing unplanned downtime was a non-negotiable criterion in the county s RFP process, and a primary reason for installing Stratus ftserver systems. At the County s primary site, two CAD applications run in virtual machines (VMs) on one ftserver system while a second ftserver system supports the Microsoft SQL CAD database. The overflow/disaster recovery site mirrors primarysite operations in near real-time and can take over complete operations in a very short period of time. The Ambulance Service of New South Wales is the third largest in the world responding to a request for assistance every 28 seconds; that s over one million calls per year. #20120805 The Kusnetzky Group 2012 Page 6
Downtime is not an option. They must be operational 24x7. Their primary data center supports four dispatch centers. Additionally, they maintain a mirrored disaster recovery site. They chose Stratus ftservers because they are field-proven to provide the highest achievable levels of uptime and are easier to manage than alternative solutions. SUMMARY If your organization requires applications to always be available and perform well but believe a clustered solution is too complex and costly, fault tolerant/continuous processing solutions should be considered. This approach is a simple solution to what can be a very complex problem. It can also reduce your organization s overall costs of hardware, software, storage and staffing while providing disaster prevention rather than offering a response to an IT solutions disaster. #20120805 The Kusnetzky Group 2012 Page 7