Announced in April, Dell's 1655MC blade server finally cleared the last of the delays in its gestation period and arrived as a (modestly equipped) $1,600 bouncing baby blade about a week ago. The time Dell expended in labor with its new offspring was not well spent. Basically, the 1655MC is a Pentium III-powered blade that comes with SCSI storage and Ethernet connectivity--an ancient configuration by today's standards, but Dell defends it nonetheless. We need only look back to other offerings from just June or July to verify its antiquity.
Apple's Xserve, for example, can be had with a 1GHz PowerPC G4 processor, 256MB of double data-rate SDRAM, a 60GB ATA/100 hard drive, a CD-ROM drive, dual Gigabit Ethernet networking ports, two USB 1.1 ports, three FireWire ports, and Mac OS X Server with a license for an unlimited number of users. Granted, its $2,999 price tag at this level shoots well beyond that of the 1655MC, but the Xserve offers an unlimited user license, and that certainly offsets some, if not all, of the hard dollar difference.
IBM's initial crop of blades, announced at about the same time as Dell's, were powered by Pentium 4 Xeons that paid not a whit of attention to the traditional blade mantra "low voltage, low power." According to Tom Jarosh, IBM's blade VP, "Our customers want blades to be real servers and they want them to be ready not just for the edge of the network applications but all the way back into the enterprise." IBM claimed to know that because it talked to its customers.
In fact, you'd have to go back to December of last year to find a Pentium III blade from a top-tier vendor (HP, whose first blade was Pentium III-based). Why did Dell come to market with nearly a year-old platform? Because Dell talked to its customer base as well.
The Pentium IIIs were part of the original plan back in April and Dell's customers declined the added expense of switching to Xeon--along with any attempt at cranking up the power requirements. Says Dell, the server targets "scale-out" applications--Web servers, Oracle or SAP application servers, thin client computing (Citrix), security (firewall/VPN), and media streaming--and using Xeons would not have resulted in significant performance improvements within that application segment, but would have raised the price.
The emphasis on price is surprising from several fronts. AMD's Athlon MP processors could have provided enhanced performance without significantly elevating the 1655MC's cost. And even if they did raise the price, it could have been offset by using IDE hard drives instead of the SCSI drives that are part of the package. But the substitution of drive types was not viable, according to Dell. Again in discussion with its customers, Dell found that there was resistance to IDE drive types--a genuine bias against them within its customer base--despite the cost savings. (Dell admitted that there was practically no difference in performance between the two drive types within the context of the applications the 1655MC addresses, but was not prepared for the arduous task of re-educating its customer base on the issue at the time.)
Granted, Dell's comparatively low-cost portables have replaced ThinkPads in several major corporate environments, often based solely on price--but that's a money arena. Dell may have misjudged the needs of this segment of the server market with the 1655MC by not looking outside of its own customer base. Even at the low end, the trend seems to be that off-the-shelf performance and scalability tend to overshadow price alone. (It's that TCO versus ROI thing again.) Dell may pull it off when it introduces the rest of its modular line next year but, right now, the bantam-weight 1655MC doesn't look like it has much promise.