Evangelizing Mainframe
Print Email

Today’s Technology Challenges Are Actually Old History

I often talk with customers and peers that simply don’t understand z/OS and its various “oddities.” They don’t understand the concepts of Job Control Language (JCL), partitioned data sets (PDS), CICS, LPAR and so on. But even if you aren’t familiar with z/OS, that last one probably rings a bell.

Have you heard about LPARs on other OSs? Most of us have, some vendors call them logical partitions, while others call them zones, virtual partitions and virtual machines. While there might be some differences, for this discussion they are all basically the same.

For those interested in history, IBM introduced the first partitioned business z/OS version in the early 1970s as well as the term virtual machine, which is still in use. Clearly, “history repeats itself.”

The business drivers for LPARs came from the need to reduce cost while increasing business agility. They allowed for the consolidation of hardware while separating workloads. Such separation was sometimes for business reasons (one application was busy at certain times of the day and others were busy at other times), security (Customer X and Customer Y data could not be intermixed), architectural (production versus test) and many others. Subdividing one large system eliminated the need for many smaller systems, and it was more cost effective. Not only was the cost of the machine reduced but also the required physical space, plumbing (most were water cooled) and other associated expenses.

For distributed systems, the recent move to virtualized systems was driven by the same factors but with a bit different historical influence. Customers are looking to reduce the costs associated with their distributed implementations, which were based on many small systems. For example, many are using racks to reduce space and air-conditioning costs, however, the business, security and architectural needs that drove these systems to be separate in the first place cannot be easily eliminated. Even more recently, when businesses are considering private and public cloud implementations, they are generally looking for these same benefits. In fact, much of the debate around moving to the cloud focuses on maintaining these key requirements not the cost saving.

Again, history is repeating itself.

Whether mainframe or distributed, systems operate independently but administration and other costs are increased. When these are combined, the shared environment is not the sum of the separate ones. The proliferation of z/OS partitions simplified many things, but it made management more complex. That is, tuning one resource on one partition impacted other partitions sharing that resource. Or when two applications wanted all of the resources at the same times, operations did not run as well. And of course, as systems got more complex, applications spanned multiple partitions.

So, in addition to having to understand the inner working of any one system, it became necessary to look beyond the boundaries of these systems and understand their interactions. Again, anyone following virtualized or cloud-based applications will have seen this same discussion about how to effectively ensure these applications perform optimally. You need visibility both within the components of the clouds and their interactions to get the complete picture. You can’t manage a cloud by simply watching CPU utilization. So if distributed systems are following in the footsteps of the mainframe, then I guess we need to ask if the mainframe solved these problems.

The answer is yes, although as the mainframe, distributed and cloud environments interact, the need for an even broader view increases. Application performance monitoring tools are uniquely positioned to understand this complex world to ensure you get the most out of it.

In closing, nobody really wants to understand JCL, but it’s just like a bat file or shell script that says what is required to run a program. A PDS is the equivalent of a folder that contains your documents and CICS is one of the first application servers. Things really are so different after all—it’s history repeating itself.

Richard Nikula is vice president of development at Nastel Technologies. He has more than 30 years of experience in IT.

Posted: 2/13/2012 6:41:58 AM by Richard Nikula

Print Email

Join Now!