Guild Companies

One of the toughest things that any IT manager or programming professional has to do is keep track of the strengths and weaknesses of the myriad computing platforms out there. All of our careers are based on making valid assumptions about what is hot and what is not, and about what is steady and what is trendy. This is one of the reasons why analysts at Gartner invented the Magic Quadrant, which is a visual representation of how the different platforms stack up against one another.

I recently talked to Gartner analyst George Weiss, one of the analysts who put together the Magic Quadrant for server platforms twice a year, about how the different server platforms are faring against each other as they increasingly get the kind of features that used to be only on expensive proprietary mainframe or midrange machines.

They say a picture is worth a thousand words, and that is why Gartner invented the Magic Quadrant. A good picture is worth money, too, if it can distill a lot of disparate information. Lots of money. That's why you'll have to go to Gartner and pay them to actually see the most recent Magic Quadrant for servers and the accompanying report that goes with it. Weiss says that the company will update the report at the Gartner Symposium in September, so you might want to wait until then.

In any event, Weiss is happy for us to tell you about where the vendors stack up in the Magic Quadrant, so we'll have to do in words what we can do with a picture. The Magic Quadrant is a square that is cut into quarters. Along the Y axis heading up the left side of the square is Gartner's reckoning of a vendor's ability to execute on its vision of server computing. Along the X axis on the bottom of the quadrant heading toward the right is Gartner's assessment of the completeness of each vendor's vision for server computing. Being at the bottom left of the quadrant is bad, and being at the top right is very good. The bottom left square is what Gartner calls niche players, the bottom right square is where visionaries hang out, the top left square is where the challengers are, and the top right is where the leaders are.

The leader in the server market, in the Magic Quadrant from early 2003, is IBM with its pSeries. Weiss says that IBM has moved into a more commanding position because of its aggressiveness in the RISC/Unix and Intel-based Windows and Linux markets. According to Weiss, the pSeries appears in more than 90 percent of the Unix shortlists that customers discuss with Gartner analysts, up from about 50 percent two years ago. This is a dramatic shift.

While the iSeries may not be in the upper right quadrant, like the pSeries, because it shares much of the same technology, the iSeries will benefit secondarily from the better position that the pSeries has against its rivals in the server space. As for the iSeries specifically, it continues to be a contender. "As has been the case with the zSeries, partitions seem to be a mid-life kicker for the iSeries," he says. "The iSeries is still up there above the contender line, and still hangs in there. It is a good enough box for a large part of the market."

For years, the AS/400 and iSeries platform has been just a little above the challenger line and a little bit inside the niche player line. There are worse places to be in the Magic Quadrant. By the way, Red Hat Linux for X86 servers is just a smidgen to the right of the iSeries. SuSE Linux on X86 is a more visionary product, as far as Gartner is concerned, but it hasn't broken above the challenger line as Red Hat has. Weiss says that, thus far, Red Hat and SuSE are the only vendors offering a compelling platform for Linux as far as commercial server customers are concerned. He also expects that the market will sustain companies that offer a viable alternative to the Microsoft Windows platform, because if there is anything that the server market--and indeed, just about any market--loathes, it is a single, dominant player.

That said, enterprises like to pick one supplier for the whole Linux stack--Web, file, print, mail, firewall, and application serving--and that means there's room for both Red Hat and SuSE, and maybe even a few other Linux distributors, in the market. One of the key things that SuSE will need to do is get certified to run Oracle's Oracle9i Real Application Clusters. Both Red Hat and SuSE are already checked out on IBM's alternative to RAC, which was announced at LinuxWorld a few weeks ago and is called DB2 Integrated Cluster Environment, or DB2 ICE. Weiss says that, in commercial computing, two-node and four-node clusters of Xeon-based servers are becoming more common, so this is not a fad. Unix clustering, Parallel Sysplex mainframe clustering, and clustering iSeries machines for high availability is common among big companies in the Global 10,000, but the kind of clustering that Oracle RAC and DB2 ICE do is different. These clusters allow a database to be distributed over many machines and to share work, not just cover each other in the event of a failed server node. By the way, OS/400 has had this capability since DB2 Multisystem was rolled out with OS/400 V3R7 in late 1995.

The joke that mainframe and AS/400 resellers used to tell is that it takes nine months to have a baby, and distributed computing using X86 servers was like having nine women have a baby in a month. They would laugh and point out that this is not possible. While they were right in their conclusion at the time, this joke was funny, but flawed. What Oracle RAC and DB2 ICE try to do is have nine women share responsibilities for having a baby in nine months. And guess what? It appears to work. This may not be funny, but finally database clustering technology has evolved enough that real distributed commercial computing is possible. (That's my assessment. I don't know what Gartner thinks about Oracle RAC or DB2 ICE.)

The Unix machines now sold by Hewlett-Packard (the HP-UX RISC/Unix machines and the Intel-based ProLiant machines from the former Compaq) used to be a little bit higher in the upper right quadrant, but have slipped as the HP-Compaq merger and its product transitions sink into the customer base. Sun Microsystems held a more commanding position in recent years, but it has lost momentum since the dot-com bubble burst, and IBM has pushed its way into the upper echelons of the Unix market alongside Sun and HP. Dell's PowerEdge servers are in the upper right quadrant, just over the visionary line, executing a little better than HP's ProLiants but not quite as visionary. Gartner reckons IBM's xSeries line is a little more visionary than HP's ProLiants, but that IBM is not executing quite as well. These Intel-based and Unix-base vendors are pretty tightly packed in the quadrant.

Gartner reckons that Unisys, with its ES7000 Wintel servers, and Fujitsu-Siemens, with its Sparc-compatible PrimePower machines, are visionaries, but says they have not yet broken into the promised land of the upper right quadrant. Fujitsu-Siemens is still a niche player in the Intel-based server space, and Sun is even more of a niche player with its X86 servers running Solaris or Linux. The Bull Escala Unix servers, which are rebranded IBM pSeries machines, are considered niche players, as are the HP AlphaServer machines running Tru64 Unix.

It will be interesting to see if either of these latter machines is even on the next Magic Quadrant. I am also curious about what effect, if any, the revamping of the iSeries will have on its standing. And with "Madison" finally proving Intel's mettle with the 64-bit Itanium processors, and HP aggressively pushing that chip across its revamped Integrity line of Itanium servers, it will be interesting to see if HP rises in the visionary and execution rankings to pull ahead of the IBM pSeries line.