IBM's zEnterprise Is A Game Changer For Application-Platform Choice

A quick note on a big announcement today by IBM that is being rolled out as I write this. No, I don't have a crystal ball - my colleague Brad Day and I spent a day in Poughkeepsie in late June for the full scoop - provided under NDA. The announcement is massive, so I'll just lay out the high points and a few of my thoughts on what it means to apps folks. I'll leave the deeper I&O/technical details to Brad and others in subsequent posts and research. My goal here is to get a conversation going here on what it may mean to apps people in your IT shops.

What's in the zEnterprise announcement?

  • It's a new computing environment that unifies Linux, AIX, and z/OS on a new server complex that includes mainframe servers, x86, and Power7 blades under a single set of management software: the zEnterprise Unified Resource Manager (URM).
  • A 10 Gb private data network joins the new z server (z196) and zBX - an ensemble that houses racks of x86 and Power7 blades. It also includes an intra-ensemble network that is physically isolated from all networks, switches, and routers - permitting removal of blade firewalls.
    • One client claims a 12-to-1 reduction in network hops by eliminating blade firewalls.
  • The z196 permits up to 96 Quad-core 5.02 ghz processors, 80 available for customer use, and 112 blades.

What is the impact on applications people and application-platform choice?

zEnterprise is a monster announcement that heralds a long laundry list of improvements - it would be impossible to cover all of the ramifications in a single blog post; however, a brief glimpse of some of the most notable improvements that affect applications folks include (zEnterprise as compared to z10):

  • For Linux - up to 60% performance improvements and 35% cost reduction
  • For z/OS - up to 40% performance improvements and 60% cost reduction
  • For XML - up to 30% to 50% performance improvements
  • For VSAM - up to 40% performance improvements

There are literally hundreds of improvements in the announcement over z10 - memory improvements, speed improvements, and power consumption improvements - but I keep coming back to the union of traditional z with blades via private data networks - all under URM (manageability). zEnterprise will force folks to reconsider their application platform strategies. 

Many will ask themselves: "Why would I choose to:

  • Remain embroiled in the religious wars around which vendor's platforms are best suited to run this workload versus that workload
  • Manually configure and manage blade instances
  • Pay for redundant management capabilities

. . . when I can choose between zOS, Linux on System z, or AIX/Linux on blades from a single vendor and manage it under a single resource manager - greatly simplifying my computing environment?" I'm not saying there may not be reasons to pause - too deep an investment in a single vendor comes to mind - but one could argue that most vendors are now fusing software and hardware stacks to leverage performance and manageability. Thinking forward to what happens when Tivoli, Rational, IGS, and other IBM groups outline how zEnterprise shapes their product strategies brings some really interesting capabilities to mind.

I'm hoping all of you will jump in on the pros and cons for apps folks of the zEnterprise announcement and generate a healthy debate. Please also see the full announcement at: http://www-03.ibm.com/systems/z/news/announcement/20100722_annc.html?cmp=zenterprise&cr=ibm&cm=S&csr=systemshp

I've only scratched the surface of the full announcement here, and from a largely "What does it mean to apps dev and delivery?" point of view. The I&O folks will surely have much more to discuss, at technical levels far deeper than concern the apps development and delivery role, and some of my colleagues - Brad Day, J. P. Garbani, Glenn O'Donnell, James Staten, and Galen Schreck - may opine on those aspects.

Comments

Fabulous announcement

As a former mainframe systems programmer now turned web developer, I think it's high time that IBM makes it far easier for the enterprise to be able to migrate their applications into the modern age. Yes, they have been going there, but this looks like it might be the [big] step that will make it all happen. Hurray!

[Gosh, does this mean that Cobol may actually die now?]

zEnterprise

Hi Nancy,
Thanks for your comments. I see a huge need to de-mystify the hopelessly complex application and systems environments that have built up since the days of monolithic computing, and zEnterprise (as you point out) as a welcome step in that direction.

COBOL is as dead as the mainframe (meaning it is alive, well and healthy) - it will (IMHO) live on well into 2025 and beyond.

Indeed

There is no server that I know of that has the I/O capacity of a mainframe, so the big boxes have lots of usefulness for as far as I can see into the future. This is an unfortunate design choice made by Intel way back when, and I have seen nothing that indicates that it is really being addressed.

It would be nice if COBOL would die, but the fact is that it is a relatively efficient language at run time; it is just lacking in the programming side. On the other side, we have all these programmer-efficient languages, but they suck at run-time, especially interpretive languages (like my favorite, PHP). What's a poor CTO to do?

zEnterprise

>>> What's a poor CTO to do?

I think we in IT can take a lesson from US auto mechanics - who once worked on American cars only, and needed only SAE wrenches. When European and Japanese autos became mainstream in the US, they didn't throw away all their SAE wrenches, they bought metric wrenches, and now use the most appropriate tool for the job.

Many programmers see new technology as absolute replacements, rather than another tool to use. Now it's fair to say that some tools SHOULD be retired - the earliest forms of BASIC and 4GLs for example. But there will always be that one shop that specializes in working on "vintage" automobiles - so even the "shed this obsolete language" mandate isn't true 100% of the time - use your tools until they are no longer of any use. Balance it with "How many tools should I retain / can I manage?" to avoid your tool chests from becoming hopelessly and needlessly complex.

Sorry for mixed metaphors, but you get the idea.

Dissenting opinions?

Hi, In hyderabad we have

Hi,

In hyderabad we have different institutes for training of IBM Mainframes but if you want to go for good institue i will suggest for reliance global services. it is located in ameerpet lane opp to chandana brothers. for more details call them on 040-66666165/156. they are authorised training partners of QAI. so they are providing software testing with certifications,

IBM mainframes training in hyderabad|Bangalore|Chennai