Test bench
Will the server support a good variety of operating systems, applications, and connections?
Futureproofing
Does the server provide you with room to expand and have redundant or swappable components?
ROI
The age-old comparison of price, performance, and features.
Service
What warranties and service contracts are available? Can you get prompt service at a reasonable price?
How we tested
We ran Ziff-Davis Web Bench on these Xeon based servers. Unfortunately with the relatively new Microsoft 2003 Server Operating system that some of the units were pre-installed with there were a few minor glitches and we could not get a reliable result from the testing. Therefore we made the decision to re-format each of the machines and install 2000 Advanced Server.
Also we initially had the intention to run Web Bench out of interest to see how the different Xeon machines stacked up against the -next generation" of machines that we tested in the December edition of the magazine (Opteron, Itanium, UltraSparc & G5). However, due to the fact that between these two series of tests we needed to change the test rig client machines quite dramatically, and amongst other things some of these Xeons had been used to act as clients to test the other servers as featured in the December review. And now these Xeons were the target machines so we could not keep them in the rig, therefore the scores unfortunately will not be comparable.
The test rig comprised of a dedicated controller system, the target server on test, a Dell Gigabit switch and the remaining network ports populated with client systems, each of which is capable of running multiple virtual clients.
The Dell blitzed the field with a maximum of 1923 simultaneous requests per second (RPS) on the 60-client load. The Hallmark while only a dual CPU configured machine showed a decent result of 1772 RPS at the 60-client load. The HP and Xenon were very close to each other scoring around the 1795/6 RPS with 60-clients. With the HP very slightly edging out the Xenon overall though.
As you can see from the results achieved HTTP requests are not very CPU intensive, the dual CPU machine was almost as quick as some of the quad boxes, the performance differences are mainly down to system memory, disk/network I/O and incidental performance tweaks by the manufacturers. These machines are all designed to do much more than simple Web serving, however the results of this test give a good comparison of the underlying system performance regardless of the number/speed of the processor and goes to show that MHz are not necessarily the be all and end all to server performance. And it also goes to show that servers should be matched with applications and forecasted server loads.
Company: Joanne's Pies. This business wants to install some mid-level servers for database and e-mail use.
Approximate budget:Â Open.
Requires:Â Six servers capable of supporting up to four processors, with two processors installed.
Concerns: With the expected growth over the next few years, the company wants the option to install additional processors and memory. With storage and all other aspects, redundancy will be very highly regarded.
Best solution: Choosing the best server comes down to a process of elimination. The Hallmark can only fit two processors, while the HP's 2U chassis provides for very limited expansion making it more suited to a high-capacity data centre. Given the Dell's vast size and enormous expansion capability, which is probably too much for this particular scenario, that leaves the XENON as the perfect fit.
Dell PowerEdge 6650 and HP Proliant DL560
The HP houses four Xeon CPUs in a tiny 2U chassis and still supports two SCSI drives and redundant power supplies. The HP would be perfectly suited for a SAN environment with the storage external from the servers, or a data centre environment where the monthly cost is based on a per unit of rack space. The Dell--on the other end of the scale--also has four CPUs, as well as a full hotswap drive array and plenty of internal expansionno wonder it's a gargantuan 7U in height. This gives it overall great expansion capabilities.
There is surely a Xeon-based system in this review that would suit most corporate server applications and budgets. A distinct move has now been made away from dedicated server internal storage towards support for external storage arrays. This is particularly evident in the very powerful and very small 2U rackmount based quad Xeon system that HP supplied, enabling you to fit twice as many CPUs in the same rack space as that occupied by either the Hallmark or the XENON.
Hallmark, Xenon? Why include these two almost unknown brands, yet exclude IBM who is one of the largest x86 server manufacturers?