IBM unleashes a Linux T-Rex

What costs three-quarters of a million dollars and comes with a one-year warranty? Why, IBM’s eServer zSeries 990 mainframe, of course.

But after years of dealing with Big Blue, I’ve learned to shake my head in bemusement at some of the kooky things the company does while I admire the amazing technology it keeps pumping out.

IBM’s z990, code named T-Rex, was built by IBM from the silicon up. It was introduced in May as a follow-on to the z900, but this baby is more than just another mainframe. 

A clustered z990 can handle about 13 billion transactions a day. The secret to its success will be that it runs Linux as well as its own z/OS and the virtualization software can create thousands of Linux servers in a box the size of a refrigerator.

The CPUs can sense changes to levels of work flow and respond by delivering added capacity, something I am told is unique to the z990. Because IBM is IBM, it tends to have a unified strategy from the mainframe on down. The z990 is part of IBM’s On Demand technology that includes autonomic features such as adding capacity when needed. A standard 16-way system with four redundant processors facilitates self-healing, and half the code in the operating system is there to prevent failure, according to Peter McCaffrey, an executive with the eServer zSeries division.

It is true that Linux servers are cheap. Many companies don’t even do capacity planning anymore; they just buy another six pack. It costs less to buy additional servers than it does to figure out if you need them, according to John Parkinson, chief technologist at Cap Gemini Ernst & Young. But the fact is, IBM is seeing some very large Fortune 500-type customers who are tired of rack after rack of servers.

A mainframe solution like this shines if you are using stand-alone Linux servers poorly.By consolidating onto a single box, virtual Linux servers can be carved up in smaller systems with just a single entity to manage.

IBM says if you have 1,000 servers, you are probably having trouble with at least 25 of them. For every 500 servers, IT can create a partition inside the z990. If a partition blows up, the work gets redirected to other partitions. The mean time between failures is rated at 50 years.

I spoke with Joe Poole, a Reading, Pa.-based technical support manager at Boscov’s Department Store, a privately held retail chain with 40 stores and $1 billion in sales last year. Boscov’s bought the zSeries when its old mainframe ran out of capacity.

Poole likes his z900 running z/Linux because he no longer pays for the AC or electricity, gains back the floor space, and doesn’t pay maintenance to the hardware vendor. Plus, the disaster-recovery plan becomes much easier. And instead of hiring another administrator as the server count continues to climb, Poole just makes sure his Windows administrators are cross-trained on the mainframe.

Of course, Parkinson and Poole both say it doesn’t pay to buy a mainframe unless you can make a clear-cut case for consolidating hundreds or even thousands of servers. But if you find yourself in that situation, consider the z990, and don’t worry too much about the one-year warranty.