Solutions

Ordering additional mainframe hardware was once a regular, accepted part of the budget cycle. This process made capacity planning a far less challenging task than it is today. Performance problems, regardless of the cause, were easily addressed by adding more hardware. Performance analysts and capacity planners were able to deal with performance issues with little concern about the cost.

Economic uncertainty and recent world events have changed this paradigm. Now that every dollar in the information systems budget must yield a maximum return on investment, hardware upgrades are delayed as long as possible. In today's world, simply adding new hardware is not the most efficient or cost-effective way to manage performance problems. At the same time, reductions in staff from downsizing and the retirement of experienced mainframe technicians are causing mainframe technical expertise to diminish. These issues make performance management that much more of a challenge.

The demand for continuous systems availability and reliability is increasing exponentially. What was once a reasonably controlled user population has expanded to everyone with an Internet connection. Web-enabled legacy applications are causing transaction volumes to explode, putting a greater strain on IT resources.

"Do more with less," is the mantra, but what is the best way to accomplish this while providing the required service and performance? While hardware costs are dropping, software and people costs are increasing. As the total cost of ownership (TCO) rises, each business transaction becomes more costly. One of your many challenges is to control costs while meeting service level objectives. In today's world, simply adding new hardware is not the most efficient or cost-effective way to manage performance problems. Contrastingly, TCO can be best reduced by optimizing your existing resources, improving application performance, and deferring costly CPU upgrades.

The Old Ways Aren't Enough
The traditional methods of dealing with performance issues are seldom adequate in today's environment. Many system programmers and performance specialists have learned to work around performance issues.

Not too long ago, well-defined batch and online processing windows made it possible to change processing times in order to take advantage of well-known periods of low activity (valleys), where resources were more plentiful. Today, while batch is still a key workload, online processing occurs 24/7, turning the picture of yesterday's peaks and valleys into a plateau of near-constant demand. Online applications are the priority workloads day and night. Deferring work is not an option, and moving it can be a risky proposition without a way to test the impact.

Because of this, many companies looked to migrate work to distributed systems (DS) such a UNIX and Windows, but the costs of rewriting applications often proved to be prohibitory. In addition, three-tier environments were heralded as the "next new thing," but many companies became aware of the lack of cross-system expertise to manage the enterprise.

Adding to the challenge, hardware upgrades and tweaking system parameters often resulted in smaller performance improvements than expected, considering the outlay of time and money. In many shops, more than half of performance problems originated from inefficient application design and, with pressing business deadlines, programmers are forced to make it work, rather than make it work well, allowing for errors.

If optimization and tuning opportunities are ignored during the development cycle, you will pay for it later - in time, people, dollars, or an application's inability to scale. No matter how much CPU or system timing is done, inefficient applications place additional demand on the system. Industry analysts have demonstrated that it is 10 times more costly to resolve a performance problem in production than during development and testing. Time and again, these performance problems translate into lost business opportunities.

The New Ways Exist
The mainframe environment is dynamic, with daily changes for maintenance and new development. The ability to tune a complex batch window or to manage a high-demand CICS system is rapidly becoming a lost art. The right tools are essential to manage these dynamic and complex environments. The old manual tuning and optimization processes that worked so well in the past are simply not adequate to meet the demand and data volumes that exist on today's systems.

To address the changing environment, companies must leverage performance and capacity management solutions that enable you to get the best results from existing resources. These automated solutions should:

• Model performance and plan for growth
• Manage application quality
• Optimize batch processing
• Optimize CICS processing

Model Performance and Plan for Growth
To reduce costs and process data efficiently, identify targets - workloads that use a large amount of costly resources - without putting excessive artificial loads on the system. To do so, companies should implement a performance management solution that allows IT managers to track work down to the individual address space and drill down to find candidates for resource optimization. In addition, enabling such a solution will allow users to analyze CICS, IMS, DB2, and MQ transactions.

After you have identified candidates for resource optimization, it is important to test tuning options and moving workloads (or parts of workloads), and changes in the transaction mix, to ensure that production response time and turnaround remains within agreed limits.

With these tools, it is easy to test a myriad of solutions and select the best price/performer. If the hardware costs of disaster recovery (DR) are of concern, DR strategies can be tested to ensure that acceptable performance can be achieved in a variety of situations. Though the CPU impact can generally be assessed with a spreadsheet, the impact on throughput and response times requires an understanding of queuing theory (which is the core of analytic modeling).

A common question that is posed to capacity planners is: "How much will this new application cost when everyone is using it?" Users can answer this question by modeling volume changes down to the individual address space. This process demonstrates the cost of maintaining acceptable performance at the new volumes. Knowing exactly how much hardware is needed, and when it's needed, simplifies the budget process.

Application Tuning
Applications and systems are more complex now than ever before. Specialization has become the norm. Application developers and systems programmers can have an impact on performance, but they do not focus explicitly on application performance. Without an automated way to manage application quality, a shop might never recognize a poorly designed application that consumes excessive resources yet meets acceptable service levels as a performance improvement opportunity.

The demand for quick time-to-market coding forces developers to push code into production too quickly. Time is limited for adequate design analysis and testing, and other factors, like high-level languages, further complicate the environment in which they run. Due to the lack of attention to application tuning, it is often difficult for an application programmer to know which code structures will result in less efficient processing.

Until recently, the goal of improving performance through application tuning was not considered critical - to reiterate, new features and increased functionality were the goals. In reality, many application-tuning opportunities relate to problems that were introduced when the application was coded. Therefore, application tuning is a significant opportunity for large cost savings.

Application quality management (AQM), a methodology for proactively optimizing mainframe application performance throughout the application life cycle, automatically targets candidates for performance analysis while prioritizing opportunities for analysis. The AQM process provides automated application measurement, automated and targeted performance diagnosis, and prioritizing performance analysis, which results in significant IT savings through deferred upgrades and resource optimization.

Manual tuning procedures are time consuming and inefficient, and few organizations have the luxury to operate this way anymore. By using this process in the development cycle and automating the application tuning process, you can avoid performance disasters.

Add Your Feedback

In order to post a comment you need to be registered and logged in.

Register | Log in

Please wait while we process your request...