FREE email newsletters 
Three Strategies for Controlling MIPS Growth in Mainframe SOA
Simply SOA
Information Management Online, October 25, 2007
There is no doubt that demand for service-oriented architecture (SOA) is accelerating. Organizations are realizing the noteworthy enterprise benefits an SOA can bring to an organization. It is flexible; it is dynamic; and, most of all, it is realistic and pragmatic.
Advertisement
When it comes to building and deploying mainframe SOA and Web services, companies are actively seeking approaches to service-enable key applications and data for both reuse of assets and investments, as well as supporting new business opportunities. The possibilities are endless, but as we add work to the mainframe, million instructions per second (MIPS) growth remains a primary concern.
The truth is the mainframe still holds critical applications and enormous amounts of data for many organizations. While some companies are still wondering what they really want to do with the mainframe, there are immediate and impactful benefits to including the mainframe into SOA plans. There is some hesitance regarding adding additional workload to the mainframe, largely due to the concern that mainframe SOA could spiral into a myriad of detrimental software upgrade costs.
For many organizations, the mainframe has become a practical entity within their SOA providing steadfast reliability and performance, despite its shortcoming of not being perceived as flexible, functional or as cost-effective as today’s hyper-distributed platforms.
While cost concerns remain a primary driver in deciding not to include the mainframe in an SOA, consider the vast amounts of data and applications currently in use and the costs associated with not including them into the SOA. There is just too much to be gained when it comes to mainframe participation in an SOA. So, how is it possible to control MIPS growth? Here are three strategies that can help.
Stay Away from Code Generation
Mainframe SOA solutions provide an effective bridge between the Web service world and the mainframe world. Solutions that generate code for the mainframe are perceived as highly performant and efficient, but they limit options when it comes to controlling MIPS growth.
Primarily, generated code is targeted at specific environments, platforms and operating system (OS) versions. This all but ensures that easily shifting workload within the mainframe, or to a distributed platform, is not an option. Additionally, noncode generation solutions have proven effective and efficient in many companies.
Additionally, complexity affects all aspects of the infrastructure. When you are talking about code generation for solutions that essentially bridge between the distributed and mainframe environments, you’re exponentially increasing the layers and effort to build, deploy and optimize your solution.
Instead of code generation, consider an XML-based runtime solution that will enable you to shift SOA-related workload to a specialty processor or off the mainframe altogether. Additionally, IBM has announced their XML parser will take advantage of specialty engines, providing even more reason to consider an XML-based runtime.
Leverage Mainframe Specialty Processors
zIIP, zAAP and IFL specialty engines enable companies to offload work from their general purpose processor to work that doesn’t impact MIPS. zAAP is intended for Java and XML. zIIP is intended for database two workload and IFL handles zLinux. Consider a service-enablement solution that can employ the usage of specialty engines that can inherit SOA-based workload as well as leveraging IBM’s XML processor and or even deploying workload on zLinux.
It is important to note that there is some confusion regarding how much workload is going to be handled by the specialty engines. Additionally, ensure you consider what version of zOS you are currently running and the cost and time associated with upgrading and enabling the engines.
Distribute Some or All the Work Off the Mainframe
The most sensible approach for controlling MIPS growth is, of course, not to add it in the first place. This is particularly viable if there is one-to-one mapping between the required Web services and the transactions on the mainframe.
However, you can still create composite services that are not hosted on the mainframe. For example, if you are leveraging infrastructure managed services-based transactions, there is very little additional overhead required to go back and forth between a distributed platform or zLinux and your IMS-based transactions.
In customer information control system (CICS), it is typically better to have multiple steps or interactions within the scope of a single CICS transaction. In this scenario, consider just moving the simple object access protocol processing off the mainframe to a distributed hardware platform or XML hardware acceleration device. Taking this step may provide you all the efficiency you require.
While it is inevitable that MIPS will increase as you do more with your mainframe, there are practical strategies to control costs and manage workloads. By leveraging specialty engines, distributing work and steering clear of code generation, companies can enhance the performance of their existing technologies and introduce new services without incurring an exorbitant price tag. Whatever the case, seek out companies that have been down the road already – you’ll see the benefits of service-enabling the mainframe far outweigh the costs and risks of abandoning it altogether.
Robert Morris, GT Software chief strategy officer, is responsible for the planning, integration and marketing of GT Software product solutions to the global market. Morris has an extensive background in application development and integration including experience with CASE methodologies, distributed systems as well as midrange and mainframe environments. He speaks frequently at customer and industry events including Gartner Symposium, Java One, IBM Common, IBM Transaction and Messaging, and IBM CICS and IMS. You can reach him at rmorris@gtsoftware.com.
For more information on related topics, visit the following channels:
