LAS VEGAS – Ralph Crosby, chief technology officer for mainframe service management at BMC Software Inc., spoke...
to attendees at the Gartner Data Center conference on Tuesday about IT automation, concluding that the mainframe is just another big server. SearchDataCenter.com spoke with Crosby about mainframe silos, connecting business to IT on big iron, and how management of the mainframe and management of distributed systems is slowly converging.
Can you talk about the islands of automation on the mainframe: What are they, and how did they get there?
Ralph Crosby: The islands tend to revolve around where we came from. If you look at the origins of the mainframe world, you had companies that did job scheduling, you had companies that did databases. It's really in just the last few years that IBM has started to look at these things across the environment, and at BMC, we've been working on how to integrate these piece parts. So now you're finally in an environment where you have a few companies with enough visibility across all the islands. They can start saying, here's what we can do to pull those together.
Is it easier to connect business needs to IT services on the mainframe than it is on other platforms?
Crosby: I don't think it's necessarily any easier; it's just different. A lot of it is pretty nebulous because you've got this huge piece of iron that is serving 27 different applications. When you've got one Unix box and this is my accounts payable database server, and I know that box is associated with that application, the alignment becomes pretty straightforward.
What I think is happening is you're seeing crossover strategy from either side. You're seeing the distributed world becoming virtualized. They're going to run into all the same issues trying to figure out how to apportion this stuff. The other thing that's happening on the mainframe, though, is that because the size of the environments has increased, people are tending to segregate. Instead of having one big production DB2 and everything is in that one pot, you now see people putting different applications in different DB2 subsystems. So there's more segregation happening. So I think the problems are no less daunting on either side. They're just a little bit different, but I think they're converging as well.
How is asset management becoming more of an issue on the mainframe?
Crosby: The crux of asset management has to do with managing cost-effectively and optimization of your environment. If I know what my assets are, I'm making sure things are really used. Say I've got six zSeries boxes sitting on the floor. Do I need to license every product on every box, or if I move my LPARs [logical partitions] around a little, adjust my Sysplex environments, maybe put all the DB2 here and all the CICS there, can I manage my asset picture a whole lot better? When it was all sitting on one box, it really didn't matter, you had to license it for one box; have a nice day. But now that you have multiple boxes, multiple LPARs, multiple Sysplexes, customers have a lot more choices on how they swizzle this stuff.
In the session, you said that you think of the mainframe as just another big server. Why is it important that data centers think of their mainframes in that way?
Crosby: Because that's how the business thinks of it. It's a line item on a budget invoice. If the mainframe is different, you get into the thing of "Well, why do I need this different thing? Why can't I put that on my commodity x86 or Unix environments?"
Nobody has an application that just sits on one platform anymore. Your applications cut horizontally across your entire IT staff, so you've got to be able to look at it from that perspective. So there's a component of it sitting on the mainframe, but it's just part of the stack that supports the applications.
You also said that you see automation as a never-ending adventure. What do you mean by that?
Crosby: I use the arms-race analogy, and I think it's true with automation. OK, so I automate something. Even if it's heuristic and self-managing, you're going to have new requirements and new environments. You're ending up with these complex interrelationships that are constantly evolving. The automation has to evolve with it. Say you go in and start taking advantage of Software as a Service somewhere, so now I've got my transaction on the mainframe. I do a stored procedure, and it ends up on Amazon.com. How do I automate that? So the bar is always going to move.