Insights into server consolidation onto the virtualised mainframe

Written By:
Content Copyright © 2011 Bloor. All Rights Reserved.
Also posted on: The Norfolk Punt

A continuing background story for me is the legitimisation of the mainframe as just another enterprise server. But, it’s an enterprise server that does all the “innovative” things the Wintel platform is just coming to terms with (virtualisation, multicore parallel processing, “big data”) superlatively well—and has done so for ages. Even the mainframe user interface is being dragged into the 21st century by tools like, for example, CA Technologies’ Chorus.

Interestingly, perhaps the best case study here comes from IBM “drinking its own champagne”, consolidating its own internal production servers with its Enterprise Computing Model (see here for the initial vision from 2007; the current picture is similar and the vision appears to be on track).

I don’t think that this is a case of marketing driving IBM’s strategy; I think IBM really had serious problems with server sprawl. John Adams (ECM and Cloud Implementation Leader, IBM CHQ, Enterprise on Demand) now says that it has consolidated over 5,000 servers today (it’s migrating some 2,000 servers/year). He claims energy savings of over 20k MWh (enough for a small town) and a reduction in floor space of some 47,000 square feet (this is space handed back to the landlord not space redeployed for other uses).

Now, IBM is reporting the insights it has gained from its server “consolidation through virtualisation” program, which, it claims, is enabling growth without extra management overheads and providing more consistent performance. IBM’s preferred platform is now, apparently, zLinux on a virtualised mainframe and it is consolidating all of its internal production processing from 377 locations into 3 “strategic growth sites”, and just under 50 “strategic special purpose sites”. It’s currently over 30% through its consolidation journey and is measuring very significant cost savings as a result.

Of course, this doesn’t come without significant initial investment in selling consolidation to the business (addressing a significant business pain point as an early delivery from the consolidation initiative is recommended) and analysing workloads. IBM’s internal workloads, as an IT company, suit zLinux well and it is also finding considerable savings in software licencing on the consolidated platforms (although software licence savings are somewhat less with Wintel consolidations, it says, partly because you need to add VMWare licensing).

Top management vision and commitment is important. In IBM’s case, the message around fewer, larger, servers and more efficient licensing must have caused some concern to those people in IBM selling commodity servers and those selling software licences—server consolidation is a service IBM is selling to customers outside of IBM—and the first stage must have been achieving real ground-level buy-in to the project. What this really means, of course, is that hardware and software vendors need to promote innovative thinking internally, so that they can continually enhance their offerings as technology evolves. As older offerings become more affordable through improved technology and margins decrease, new high-value offerings (such as Watson, in IBM’s case) must be delivered to replace the lost revenue stream.

Any change, however good, produces both losers and winners and if it is to proceed smoothly, management has to convince the apparent losers that the new way brings them benefits too. As an aside, it always seems to me that a lot of software companies have done rather well out of selling “shelfware” and now that they are tightening up on software licensing, on one hand, and their customers are introducing software asset management to bring software under control, on the other, there is some risk to many large software vendors’ revenues! I’m not sure that some vendors contemplating turning pirated software into licence revenue have always thought through the possible unintended consequences—such as a move towards Open Source software (not free, but with less pressure for constant upgrades) and, possibly, even a move towards virtualised mainframe platforms that can make managing software easier.

Anyway, this leads into the first key lesson for server consolidation: it’s important to do your homework first and to have sufficient top management commitment behind the initiative to provide resources for workload analysis and for achieving buy-in from the people affected during the initial “ramp up”. As usual, cutting corners early on can put eventual success at risk.

The second key lesson, according to Adams, is that a significant additional return on your server consolidation investment can come from handing real-estate back to the landlord, if this is feasible. It’s a journey; you start off by targeting particular applications or technologies, or perhaps the testing environment (which has been very successful for IBM), but this can lead to a “Swiss cheese” data centre—it’s still there as an overhead but now has a lot of holes where servers have been consolidated. This is probably a necessary stage in the consolidation journey (delivering immediate benefits) but as your consolidation effort matures it culminates in a location focus—a team concentrates on a particular location and consolidates and moves all of its IT into a set of strategic locations, so that the existing location can be closed down or re-purposed. Consolidation becomes rather easy to sell if the bottom line includes not paying rent and utilities bills for significant floorspace.

Adams’ third key lesson is that, as your consolidation matures, and if you have appropriate management commitment/vision, a large part of the consolidation ROI comes from changing behaviours. For example, in a virtualised development/test environment, easy provisioning/de-provisioning results in developers handing back server resources to the pool more promptly, leading to more efficient utilisation of resourcers.

And a fourth lesson is that, as the initiative matures and people start to think about consolidation holistically, capacity planning becomes a global thing and a single business case can support consolidation generally. This helps to reduce the barriers to further consolidation.

To summarise, I was quite impressed by Adams’ story around IBM’s virtualisation experience. Overall, perhaps, the one thing that impressed me most was a diagram showing IBM’s “strategic growth sites” and “strategic special purpose” sites going forwards. This, to me, represented IT in a very large globalised company in proper control by the business, and exemplified a fine example of what I’d see as “good IT governance” delivering a business outcome. Which is, ultimately, what server consolidation should be all about….