![]() |
|
| F E A T U R E | |
Sizing Up the Quad Squad September 20, 1999 By Dave Fetters It wasn't all that long ago when using the words Intel and enterprise in the same sentence evoked the kind of hysterical laughter reserved for Happy Gilmore or Caddyshack. Many of us still find the idea slightly amusing. However, one thing is certain: Intel-based servers are gaining market share, and unless Advanced Micro Devices can pull a rabbit out of its hat with the Athlon K7, Intel servers will continue to do so for the foreseeable future. Given servers' key role in enterprise computing, we set out to round up and assess the latest, greatest Intel quad-processor servers. We invited Compaq Computer Corp., Dell Computer, Gateway, Hewlett-Packard Co., Hitachi, IBM Corp. and Toshiba America to participate in our tests, but it was no easy task getting our hands on the hardware. We finally received servers from Compaq, HP and IBM. Dell declined to participate, saying it was between server revisions. Gateway claimed similar circumstances, and added that it was in the midst of refocusing its server strategy. Meanwhile, Toshiba was unable to provide a server that met our testing criteria: four 500-MHz Pentium III Xeons with 2 MB of L2 cache, 2 GB of RAM, a minimum of 70 GB of local storage configured to RAID 5 and a 10/100 NIC. Hitachi was unable to ship a server in time for our tests. We did not specify price so the vendors could decide for themselves which model best reflected their most versatile server. Similarly, we did not indicate whether the servers should have a rack-mount or tower chassis. We put the three servers through the grinder and then some. Network Computing worked with KeyLabs, an independent testing facility in Lindon, Utah, to create our server-test environment. Because performance is paramount in servers, we spent countless hours setting up a highly challenging test bed designed to push each server past its maximum threshold. Using Bluecurve's Dynameasure 2.0, we constructed an enterprise messaging environment with Microsoft Corp.'s Exchange Server running on each tested server as the back end. We used Outlook 97 for the front-end client messaging. As important as performance is in judging these servers, we also put a considerable amount of weight on manageability, which is a growing area of concern for many IT professionals. More and more buying decisions rest on a server's ability to be managed both locally and remotely. Luckily, vendors have taken heed and made tremendous leaps in this area. Most server vendors offer integrated Web management, advanced hardware monitoring and server cluster management. Some have gone so far as to enable their management packages to be integrated with those of their competitors. While testing these servers we also took a close look at maintenance, service, scalability and relative value. We evaluated scalability from a performance standpoint and in regard to expanded peripherals. IBM's Netfinity 5500 M20 won our Editor's Choice award; the well-constructed server did the best job blending performance, management and price, and its components were top-notch. It wasn't the fastest, nor did it have the best management package, but its results were close to the high mark in those areas set by Compaq's entry, the ProLiant 6500. The Compaq server finished a close second, having demonstrated exemplary performance and scalability. Our tests answered what many people have wondered, namely that its performance lives up to its name. Blame its second-place finish on some work that remains to be done on the management side to make the server a bit more user- and cluster-friendly. Nudged into third place by a hair, the HP NetServer LH4 displayed excellent management capabilities and good performance. With such a narrow margin separating the three contenders, it's a good bet that if we added an upgraded RAID controller to the LH4, it would have taken first place. In fact, had we weighted the categories differently, any one of the three might have emerged with the top overall grade in our report card (see above).
| |
|
PAGE: 1 I 2 I 3 I 4 I 5 I 6 I 7 I NEXT PAGE |
|












