IBM Ups the Ante

IBM this week released a new generation of servers this week dubbed eX5. Now generally this type of news does NOT make my blog. After all technology companies are constantly coming out with the "next thing" and tempting their customers with new faster goodies. What makes this different (at least to me) is that one of things I have always liked about IBM has gotten better.

So let me start by clarifying my position in the server market. They are servers (blocks) and its about what you do with them. Overall everyone makes a fairly similar server, the differences between vendors are generally in the area of support and compatibility testing. Now this can be GROSS UNTRUE - I know, I am trying to keep this blog simple and relatively short; so forgive the simplification. By the way, almost every IT professional I know has at one point or another in their career had a server vendor horror story - and that eliminates that vendor from their purchasing choices for the rest of their lives (unless forced).

Anyone who has worked with me knows that I like to design solutions that create an established base of hardware and upgrade that base later for scalability. This practice has served me well for most of my career, and I dont see changing. What I do see happening is that scale factor has always been determined by hardware manufacturers limits in their server systems. For anyone who has followed IBM's system X portfolio, one of their distinguishing factors has always been that they design two different types of server systems - ones made for scale out folks like me, and ones that are designed around scaling up (single servers made from building blocks that become very large computing engines). So while you could have a 2 or 4 way blade, you could also have a 16 socket system by simply cabling building blocks together in supported models.

So while I generally didn't exercise the option for the stackable systems (due to cost per module), it was there for those customers who really needed that sort of power for their applications.

So that brings me to IBM's new generation of servers. IBM engineers focused on the 2 and 4 socket rack mounts, and the 2 and 4 socket blade models - which to me is the bread and butter of the x86 server market. They took a look at what limits virtualization scalability in servers today (memory memory memory) and released systems that have generous memory slots WITH the option of essentially doubling them with an add on blade or module. What could YOU do with a 4 socket system that has 128 DIMM slots? For IBM is simple, the customer can buy lower cost memory and use the additional slots to get to bigger total memory configurations then are possible today (at the same price point), or make mega hosts for desktop and server virtualization. Of course is you having business applications that are crying out for memory and CPU - they are going to love this server platform as well.

Now being very blade centric as a designer - I paid particular attention to the fact that the 2/4 socket configurations were actually blades being configured by chassis management to work as one machine or two. This opens the whole concept of dynamic configuration of systems on a time of day or project by project basis. In terms of Intelligent Workload Management, which seems to be the new IT (or at least the vendors pushing it would have you believe this) and Cloud Computing, what an amazing feature. With the exception of eGenera, who has been doing this awhile, one of the big limitations of virtualization has always been that you cant virtualize a single virtual machine across multiple hardware platforms. If you need more CPU and memory you would have to move the whole machine to a hardware platform that has the needed resources. IBM has started down the road of mainstreaming their technology to combine modules into a single hardware server on the fly - now that really gets the techno-lust going!

Want more information?

http://www-03.ibm.com/systems/info/x86servers/ex5/index.html