Unisen Inc. had five IBM servers running software for back-office services it provides to Canadian financial institutions. The Mississauga, Ont.-based company used multiple boxes because it was uncertain the technology was robust enough to support all its clients from one machine, says Robert Smuk,

Unisen’s president. But five machines were harder to manage and support than one, and dividing the workload interfered with allocating power to whichever client needed it most at the time.

So, when IBM introduced new on-demand processing hardware, Unisen set out to consolidate from its five servers to one machine. The consolidation is nearly complete, Smuk says, with about 95 per cent of the workload now on the new larger server. Of course business continuity remains a concern, so Unisen has a backup server ready to take over if there are problems with the primary one.

Server consolidation is getting lots of attention these days as cost-conscious IT shops strive to squeeze more from their resources.

“”The driving force behind it is trying to become more efficient,”” says Alan Freedman, research manager for infrastructure hardware at International Data Corp. (Canada) Ltd. in Toronto.

Medium and large servers are being consolidated first today, Freedman says, but the trend will increasingly affect smaller boxes as well.

Making the consolidation case

The case for consolidation has been strong for some time. Server utilization rates are often as low as 15 per cent for smaller servers based on Intel Corp. chips, says Gordon Haff, senior analyst at research firm Illuminata, Inc., in Nashua, N.H.

“”It’s almost become a cliche that server utilization is terrible,”” Haff observes. Cheap hardware has encouraged installing a new server for every new application, and hardware has gained power faster than many applications increased their processing demands.

Unisen is an exception to the utilization argument — Smuk says utilization of the new, larger system is about the same, at around 70 per cent, as utilization of the five older boxes. But he says Unisen saves money because the single server is easier to manage and more flexible.

“”People are doing this to save money, make their lives easier, improve their level of service and simply simplify what they have to deal with,”” says Chris Pratt, manager of e-services strategic initiatives at Markham, Ont.-based IBM Canada Ltd.

But consolidating multiple applications on fewer servers isn’t always easy, because several applications sharing one operating environment don’t always play nicely together. A misbehaving application can cause problems for others on the same system, and in the worst case a crashing application takes others down with it.

One reason consolidation is on the upswing is that virtual server technology offers an answer to those concerns. It insulates one application from another by running them on separate virtual machines, even supporting different operating systems on the same hardware. And even with a single physical server partitioned into two or more virtual machines, load-balancing software can still shift capacity to the application that currently needs it most, adds Steve Shaw, business development manager for the enterprise storage and servers group at Hewlett-Packard (Canada) Ltd. in Mississauga, Ont.

“”It’s really a combination of there being a recognized problem, there being a desire to solve that problem and the means being available to solve it,”” says Haff.

Where to Begin?

The first step toward successful server consolidation is to understand the current environment, says Shaw. That means analysing the servers you have and how much they are used. It’s important to consider the whole business cycle, he warns, to spot peaks in demand.

The next phase is to decide what applications are good candidates for consolidation. Those that use little of their servers’ capacity are obvious possibilities, but there are other considerations. Shaw suggests talking to software vendors to determine if their packages will run comfortably alongside other applications or what conflicts might arise.

Trying to cram too much on one server is a mistake, Haff warns. Servers can reasonably run at 50 to 75 per cent of capacity if loads are fairly constant, he says, but remember to allow for peaks.

Dan Ryan, manager of IT infrastructure and desktop management at the City of Calgary, knows all about that. A couple of years ago, the city consolidated its building permit application, geographical information system and PeopleSoft software — “”probably three of our top five applications”” — on a single server.

“”We were able to physically get it all on a single server,”” Ryan recalls. “”The downside is, we were having all kinds of peak performance problems.”” Those applications now have their own servers again, but Ryan, is now consolidating the city’s 140-plus Unix servers as part of a migration to Oracle 9i database software and Linux. Having learned from experience, he is taking care to combine applications whose processing needs peak at different times, and aiming for 30 to 40 per cent utilization — up from 10 per cent but leaving comfortable headroom.

Virtually serving and protecting

Thanks to virtual server technology’s ability to run different operating systems or versions on one physical machine, it is getting easier to bring together “”stacks”” of related applications, Shaw says. Putting together two or more applications that exchange a lot of data, for instance, can boost efficiency.

Having decided what and how to consolidate, it’s best to proceed carefully. Intrawest Corp., a Vancouver-based resort operator, recently consolidated a “”sort of spider’s Web”” of 84 Windows NT domains on around 130 servers down to about 50 domain controllers, says Philip De Connick, system architect, and soon will begin consolidating Microsoft Exchange servers. Rather than reconfigure existing servers, Intrawest is installing new hardware, allowing it to get replacement servers up and running and then migrate users. If any problems arose, De Connick says — and so far none have — it would be simple to move users back to the old systems until the trouble was fixed.

“”Like anything else, you want to do this in phases,”” advises Haff.

Shaw says solid systems management is essential with consolidated applications. Systems and network management tools — HP’s OpenView, Computer Associates International Inc.’s Unicenter and others — help monitor the systems and, among other things, warn if server utilization is getting uncomfortably high. With multiple applications depending on a server, reliability features such as redundant power supplies and hot-swappable components become even more important, Shaw adds.

Haff says some savings from consolidating servers should be invested in making those that remain more reliable.

“”If you do everything right, you could then have a more reliable system than you had before,”” he says. “”If you don’t do those things, you’re opening yourself up to more expensive failures.””

Share on LinkedIn Share with Google+