Operating Systems

There’s more than one way to save on CPU consumption. Moving workloads to System z Integrated Information Processor (zIIP) engines has been a popular option. However, many organizations are also exploring using IBM’s DB2 Analytics Accelerator (IDAA) as an option for lowering costs and optimizing the performance of the mainframe. IDAA is a specialty machine that offers faster and more predictable response times for long-running, unpredictable queries. It does this by reducing database tuning efforts and off-loading query workloads. If you’re looking into moving workloads to IDAA, here are some important factors to consider.

IDAA or zIIP?

IDAA, which runs on the IBM zEnterprise 196 or 114 with IBM Netezza 1000-3 hardware, provides benefits similar to zIIP engines. IDAA allows you to offload large, expensive, heavy-hitter queries to a separate machine, thereby decreasing consumption on the mainframe. The result is lower operating costs, improved performance and better control.

How do you know if you should move a workload to IDAA? With zIIP engines, users are dependent on the executing software to determine if processing can run on the zIIP. In addition, the zIIP might be busy, so work must remain on the expensive CPU, and the full potential of savings aren’t realized. However, at the end of the day, mainframe customers with zIIP engines can save MIPS, and that helps reduce costs.

People want choices in how to reduce mainframe costs, so think of IDAA as another option that can achieve the same benefits as zIIP engines—reducing CPU by moving workloads off the mainframe. Before you purchase IDAA, however, make sure your workloads can actually be moved to the IDAA system. That decision depends on your processing for large data warehouse applications.

IDAA can provide substantial savings because it reduces the mainframe MIPS by redirecting DB2 queries that run on the mainframe to the new IDAA machine. These queries save mainframe CPU, and application queries will run faster. This enables business decisions to be made and acted upon in a fraction of the traditional time required on the mainframe. It’s up to the DB2 Optimizer to determine if a query benefits from running off-platform with IDAA or whether it’s better to run within the DB2 system on the mainframe. Looking at the Optimizer data can tell you whether your five-page, one-hour elapsed, 20-minute CPU time query would run on IDAA. You can also see if all your queries are IDAA-worthy because the DB2 explain function contains information about accelerator usage or lack of usage. It identifies whether the query qualifies for acceleration. If not, it will give a reason code.

Monitoring IDAA

If you’re adopting IDAA, it’s important to ensure you have the right products to manage it. Look for a monitoring solution that will either help you move to the IDAA or will enable you to manage and validate that you’re getting the most benefit from the IDAA.

The solution should be able to support the IDAA through collecting statistics on systems and reporting how much savings you’re achieving versus running the jobs on the mainframe itself. The reports should include a historical view of just how well the IDAA is working for you. This is particularly important, as your organization might run millions of queries against the IDAA.

You may need to run reports nightly that summarize this data and identify just how much you're using the IDAA. By processing the system records efficiently, you can get the answers quickly. For example, a telecommunications company used mainframe monitoring technology to collect valuable performance data and mined the information to show savings of zIIPs as well as IDAA queries. So, by using the technology and understanding how processing was trending, this company was able to save millions of dollars a day.

The total cost of ownership of a mainframe is considerable. IDAA is one innovation that can help you improve mainframe performance while lowering costs. Mainframe monitoring technology for IDAA can help you achieve these objectives.