At the very least, if you were to read the press release, you'd find a very attractive product: "Built with the company's virtual partitioning technology and management software, the IBM eServer Cluster 1600 has been pre-tested for 32 IBM eServer p690 32-way systems and IBM eServer p670 16-way systems." It's even attractively priced, at least within the high-end arena, starting at just a tad below $2.5M for a Cluster 1600 with two 32-way p690 servers with up to 32 partitions, and a control workstation.
Still, I pose the question because, in case you missed it amidst the constricted economy, IBM also announced the 1600 series last November. This "second coming" is actually just the rollout of an expanded feature set--something IBM does say up front, but also something that might be easily forgotten by the time you reach the end of the eye-glazing list of new or expanded features. IBM's critics (and it seems to have collected a few detractors along the way), however, have not only read and retained every line of the press release, but they've also interpreted what they consider the subtext as well. Some of the less than glowing comments received rather sharp and deserved answers--which were provided by Barbara Butler, Product Marketing manager for the IBM pSeries.
For example, a representative for HP noted that "the Transaction Processing Council withdrew IBM's 403K TPC-C result on Regatta/p690 and removed the listing from the TPC-C Web site. As we currently understand it, the reason is that IBM could not fully meet the specifications set by TPC for a valid listing of the result. It is clear that the Regatta/p690 is still having difficulties delivering on its promises." IBM doesn't deny the claim, but does say the problem was that TPC couldn't verify the price of the "as tested" configuration, and that the p690 was restored in the latest update to the list.
A criticism was proffered that this is a "clustering for performance and scalability" design, not aimed at High Availability (HA) failover. Yes, says Butler, it is designed for clustering, and it's also focused on clustering for manageability. Most of IBM's customers at this tier are running parallel servers and HA-failover isn't an issue. Should it be or become one, IBM can supply HA software as an option for its customers with business critical applications.
HP's response: "Although IBM makes it look simple with the 'pre-configured server clusters,' in the real world the larger clusters require extensive support to get and keep them running. These clusters are very expensive, especially because of all the consulting and support costs." You could almost hear, "And your point is?" in Butler's reply. "In large and complex environments," she says, "the management and maintenance of servers is tough no matter how you cut it. Many of our customers find that, as they go forward with their clustering, they can build on the infrastructure and cut many of the costs back if not out." eLiza and IBM's focus on autonomic computing will significantly reduce costs in the years to come.
An attempt at a coup d'grace emerged in an HP comment: "IBM's PSSP (Parallel System Support Programs) software can now support up to 128 AIX images, but HP's Service Control Manager already supports 1024 nodes/images. The current PSSP version 3.4 still does not have 64-bit kernel support, which is only proof of the many steps that still have to be taken to make this into a complete solution." IBM was quick to point out that its scalability limits point to configurations that have undergone "very rigorous in-house testing" and do not necessarily reflect actual limits on the architecture. IBM does entertain bids for higher scalability systems and already has many customers who have gone beyond that stated limit. As for 64-bit kernel support, no, it doesn't exist just yet, but will debut with PSSP 3.5 in October--not such a long way away and possibility not so many steps either.
What's the point of all this? Hopefully, when you learned to drive, your instructor told you to get the big picture, not just a view of the back end of the car in front of you. It's the same here. It's as easy to brag as it is to criticize. In fact, the latter may be far easier for a vendor vying for your IT dollars. The big picture in IT is that no company would last more than a day if it sold a truly inferior product. Wading through the hollering and the hype to find the configuration that fits your application is really the tough part.
Is IBM rehashing old news to get more mileage out of it--or is HP just desperate? TalkBack or send e-mail to Bill.
