IBM Taking on Sun in Web Infrastructure?

Event Summary

May 1, 2000 - IBM, after years of taking a back seat to Sun in the Web server and infrastructure market, has decided to try to make up for lost time. CEO Lou Gerstner has given his crew their marching orders: "get out and be aggressive" against Sun.

A highly visible win for Big Blue has been Network Solutions Inc. (NSI), which recently decided to change its A.Root domain server from a Sun system to an IBM RS/6000 S80, IBM's current top-end AIX/Unix system. Although this win is more of a PR victory than a big contributor to the bottom line, IBM is (rightfully) getting as much mileage as they can from it. Ari Balogh, NSI's VP of Engineering, has been heard to say that the Sun server "ran out of gas" when trying to handle NSI's geometrically-increasing workload. As a result, IBM has followed Microsoft's lead by co-opting Sun's "the dot in dot com" slogan, stating that they are "the new dot in dotcom".

As a sidelight, IBM was "pro-active" in contacting journalists in April about delays in Sun's upcoming UltraSparc III-based systems. Sun partners confirm the delays, saying the systems should roll out in July, but they add that Sun customers are not complaining. "Sun continues to grow their business, while (IBM/Sequent) is faltering," said one integrator.

To catch part of the Linux wave, IBM is also in the process of Linux-enabling all its hardware platforms - Netfinity, AS/400, RS/6000 and S/390 servers - so all can run the same applications. Covering yet another base, IBM's Monterey project will allow AIX to run native on Intel's upcoming IA-64 platform. In the meantime, Sun has been building up its Java group, while IBM appears to have backed away from Java (and Sun, by extension) by refusing to support various Java initiatives.

Meanwhile, Sun, always happy to talk down the competition, is being "pro-active" in its own way, going after IBM and AIX in the press. Said David Gee (formerly of IBM, lured into Sun's Java group), "(Some of my former colleagues) are in a very difficult position; they're defending a declining company. The fundamental issue with IBM is that it's not gaining market share in its core business, which is hardware, and software drives hardware. Nobody is writing new applications to run on IBM hardware."

This is not a "steel cage match", but it does make for interesting reading.

Market Impact

On the hardware side, IBM has made a point of going after Sun, especially in the Web infrastructure area. First it was the RS/6000 S80 blowing the doors off of Sun's E10000 "Starfire" performance on the TPC-C benchmark, especially when price/performance is included in the equation. Then it was IBM's trumpeting the sales figures exceeding those of Sun's Starfire.

Although Sun is trying to minimize the impact (what some might call "spin control") of Network Solutions switching over to the RS/6000, this is clearly not a good sign for them. In addition, the gist of the response appears to be company-focused (IBM is "a declining company"), not product-focused. Ad hominem attacks often signal a fundamentally weak position - is that the case with Sun?

Sun and Solaris is not an inherently weak platform, in fact it is generally reliable and robust, and is still relied on by much of the Web to provide the infrastructure. However, the SPARC hardware has lost a significant part of its performance advantage (at least for transaction processing, as evidenced by the TPC benchmarks), and it's not clear that the reliability sell will last them another three years, especially as Linux gains more functionality and market presence. (We do not necessarily think AIX is a threat to Solaris, and more applications are currently written for Solaris than for AIX. But resting on one's market share laurels is a good way to let someone catch you.) When the UltraSPARC III finally ships, the performance figures should be reviewed closely.

Regarding software strategy: some have said that Sun has no viable software strategy, we don't disagree. Although Sun is spending time and money to push Java, it is not clear what the overarching goal is. This is compounded by Sun's on-again-off-again support of a Java standards group. Eventually this will all settle out, but will users be willing to wait?

User Recommendations

The small server and desktop user will have little interest in these developments, since the primary focus is on enterprise-class computing and larger Web infrastructures. Most small-to-medium users will not need the power of the S80 or E10000, and at upwards of $1 million for a system, most cannot afford them, either.

Large-scale users should compare these carefully. If you want raw transaction processing power and price/performance, the S80 is currently the better choice - as long as you have your desired application(s) written for AIX. (We are ignoring Monterey for now, since it's not really applicable here.)

If you don't need quite as much power, are in need of the applications available on Solaris, and willing to pony up the cash, then the E10000 is probably better suited to your needs. Although expensive, the E10000 is a solid system with numerous enterprise-class features.