Reducing the Cost of MIPS

Reducing the Cost of MIPS

What is/are MIPS?

MIPS is an acronym for Millions of Instructions per Second, a measurement of processing power and CPU consumption. MIPS are typically associated with running critical enterprise applications on mainframe computers.

Most hardware and software packages in the mainframe world are licensed based on the size of the machine, or the number of MIPS.

How Much Does It Cost?

The total cost of each MIPS varies between $3,000 and $5,000 per year, including all hardware and software costs. On this basis, operating a mainframe environment may consume nearly 10% to 40% of a company's IT budget, depending on the size of the mainframe landscape.

Each saved MIPS can generate tremendous savings. With an average saving of 20%-30% of an organization's MIPS usage, operating costs can be reduced by many millions of dollars, on a yearly basis.

So Why Use Mainframes?

Why not use a platform where running programs does not incur costs? 

Well, running programs costs money whichever platform you're on. Mainframe users have traditionally chosen usage fees because of the shared nature of the platform (many users, on one big computer). However, even if you run your programs in a large cluster of virtual machines, you'll be charged on the capacity you reserve and/or use. So moving from mainframes to another platform might bring cost savings, but it's not without risk.

What Contributes to the MIPS Cost?

MIPS usage is directly driven by CPU usage. There are several locations where excessive MIPS usage might be found:

1) Poor application logic:  Inefficient use of the programming language (COBOL, etc.) can cause an application to consume unnecessary resources. By improving the performance of the code it's possible to lower CPU resource consumption. 

2) Excessive use of SQL commands: The SQL language, designed to ease the development of RDBMS related logic, has the potential to create high cost queries. 

3) Inefficient data models: Building an efficient data model is an art. It requires a deep understanding of the application, and of the principles of optimal relational DB operation. 

4) Changes, in use: Structures that started out as very efficient can often become degraded, due to changes in the application logic or to the data. This can also result from the introduction of new features to the DB2 version.

Strategy for Reducing MIPS Usage

The prevailing approach is to look at each of these “resource-hogs”, sequentially. There are also various optimization products or services that monitor and point to bottlenecks in one or more of the areas mentioned above. 

Ideally, the different viewpoints should be considered in combination, to understand where most of the savings can be generated:

•    Analyze several aspects of the DB2 database.
•    Catalog the metadata and SQL syntax tables. 
•    Catalog the application access information (Plan and Statement tables). 
•    Take a comprehensive view of the actual use of each element, and the related data volumes.

Specialty Engines

The exploitation of IBM specialty engines can help make the mainframe more cost-effective when it comes to everyday transactional processing. 

Three specialty engines - Integrated Facility for Linux (IFL), System z Application Assist Processor (zAAP), and System z Integrated Information Processor (zIIP) - have been optimized to handle new workloads. These engines offload application processing from the mainframe’s General Purpose Processor (GPP), and run the target workload more effectively. 

The IFL provides substantial software savings when it comes to consolidating Linux processing on the mainframe, typically reducing processing costs by up to 100 percent.

The zAAP is designed to run new workloads such as processor-intensive, Java-based applications. It also enables XML processing.

The zIIP’s pay-off comes from data and transactional workloads driven by various online systems in a distributed systems environment. These workloads typically use massive amounts of CPU cycles.

By hosting specialized workloads, processing is excluded from the overall MIPS ratings that determine mainframe software costs. Also, any work running on these specialty engines frees up cycles on traditional processors. This stretches out software maintenance cycles, and, in turn, reduces long-term costs.

Best Practices

Here are four factors to consider in optimizing costs, and in selecting the tools to help you:

1) Analyze and tune: Analyze applications with an eye toward reducing MIPS consumption. There are tools to help you identify performance bottlenecks and inefficient code, such as poorly performing SQL statements. Performance optimization tools (e.g. DB2 tuning tools) can be used to address the problems you find, and reduce peak capacity requirements.

2) Set alerts: Mainframe monitoring tools can monitor capacity and alert you when consumption is approaching the limits you’ve set. You'll be in a position to know when you need to head off a major spike in monthly costs. 

3) zIIP it, when necessary: When evaluating tools for monitoring and managing mainframe capacity utilization, look for ones that are efficient in their own use of mainframe capacity. Ideally these tools should offload a substantial portion of their processing to System z Integrated Information Processors. 

4) Test drive: Assess in advance a tool's potential to lower your peak capacity. You might look for a vendor who can test drive tools that analyze your System Management Facility (SMF) data. Determine your peak capacity and identify what’s running during peak periods. Then leverage this information to determine the return on investment you can expect from the tools. 

Today’s technologies mean that businesses working with mainframes can make significant savings when it comes to MIPS. Ensure that your annual budget is not getting eaten up by following the suggestions about and save revenue for other business processes. In the current climate competition is high and so the business that uses the most agile practices and available tools will be the one that wins out.