X
Tech

IBM announces Opteron support but lacks business rationale

Although AMD's Opteron has not yet shipped, IBM will be one of the first major software vendors to announce support for the dual personality processor that will be able to run 32- or 64-bit operating systems and applications.
Written by David Berlind, Inactive

Although AMD's Opteron has not yet shipped, IBM will be one of the first major software vendors to announce support for the dual personality processor that will be able to run 32- or 64-bit operating systems and applications.

According to IBM spokesperson Sean Tetpon, IBM will showcase at LinuxWorld its DB2 database running on an Opteron-based system provided by a Newisys, a newcomer to an already crowded server market. The system will be running a 64-bit distribution of Linux provided by SuSE.

The decision to support Opteron is especially interesting in light of the fact that IBM's hardware division is turning a cold shoulder towards the AMD processor. It was my investigation into Opteron's chances for success that had me so captivated by IBM's decision to support it.

When asked to explain the business rationale for moving forward with Opteron support, however, Tetpon was unable to make an obvious connection between selling Opteron-enabled software and IBM's bottom line. Instead, he offered four reasons. They are the sort of reasons that most businesses could never use to justify the additional expenditure of resources in today's business climate. Since none of Tetpon's justifications included "because customers are asking for it," I was hoping for some critical insight into Opteron that hadn't yet surfaced.

The first reason was that it only took two days to do the port. "Porting" is a process that developers typically go through to take an application that runs on one platform, and make it run on another. Some applications are more portable than others and, according to Tetpon, DB2 was designed from the bottom up with portability in mind. "That's why," says Tetpon, "it only took two days." If "we're doing it because we can" is reason enough, I have a list of processors that IBM should start porting DB2 to as well.

The second reason is tangentially connected with the first. According to Tetpon, IBM prides itself on the fact that DB2 is available on more platforms than any other database. This widespread availability across platforms is in DB2's heritage, and therefore it's DB2's birthright to be ported to Opteron?

Tetpon's third reason was because Linux will run on Opteron. Tetpon says, "IBM is the biggest supporter of Linux. Therefore, since Linux runs on Opteron, it makes sense for IBM to port DB2 to Linux on Opteron." So, do it because Linux runs on Opteron and IBM likes Linux? That still seems a bit weak.

Fourth and finally, Tetpon said the DB2 engineers liked its performance. "Compared to what?" I asked. After all, without any context--say, a benchmark comparing Opteron to Intel's Itanium--one doesn't simply go out and buy something because it's fast. I race bikes. But I don't go into a vacuum to make decisions about the next frame or wheel set that I'm going to buy. I compare because I either want the fastest (usually the lightest) and am willing to spare no expense at getting it, or I want the best price/performance--the fastest that fits my budget. Yet, Tetpon was unable to offer any such performance context for this justification.

So, absent of any real business justification for offering Opteron-based versions of DB2, (or any of IBM's other software for that matter), I was left wondering if maybe there was more to the performance issue than meets the eye. At last month's TechXNY, AMD Computation Products Group marketing vice president Ed Ellett claimed that a single Opteron would outperform two of its closest cousins in the IA-64 family. In other words, for less money and fewer processors, you can get more performance out of an Opteron-based system than you could out of a similarly configured dual-processor IA-64-based system. But Ellett offered no benchmarks to prove his point.

If IBM's DB2 engineers like Opteron's performance, they had to have been comparing it to something. It's also safe to assume that those same engineers had access to the fastest Itanium 2 prototypes (perhaps those based on IBM's EXA chipset). If I had to take a wild guess, Opteron might very well be doing some slam dunks against IA-64 behind closed doors. Whether that's the case or not remains to be seen.

What do you think of IBM's decision to support Opteron? Share your thoughts in our Talkback forums or e-mail me at david.berlind@cnet.com.

Editorial standards