We performed several system tests to analyze the servers' disk I/O, Ethernet performance and Web performance (see test results). Intel's Iometer tests provide a performance validation of the hard disk subsystems by generating I/O workload and performing a variety of simulated tasks. NetIQ Corp.'s Chariot tests Ethernet NIC performance, both in send and receive mode. Caw Networks' Web Avalanche tests provide raw Web performance numbers under an untuned copy of Microsoft IIS.
In all but the Intel Iometer disk-performance tests, results were almost too close to call, with less than a 2 percent difference between the best and the worst performance. However, the Iometer tests did uncover some strengths and weaknesses. In the database test, for example, the Dell PowerEdge 6650 outperformed IBM's eServer xSeries x255 by about 25 percent across the board. The HP ProLiant DL580 G2 showed anomalous results in both the maximum I/O test--more than double the throughput of the IBM and Dell servers--and the sequential test--less than 30 percent of the throughput of its competitors. HP pinpointed the ProLiant's RAID controller as the underperformer, but was unable to explain the high maximum I/O score.
Intel Iometer Tests
We used three of the Iometer suite's basic tests--the database, maximum throughput and maximum I/O rate tests. We also customized a test. The database test used a 2-KB transfer request size with a mix of 67 percent reads and 33 percent writes. This proportion represents a typical database workload.
On the maximum throughput test, we used a 64-Kbps transfer request, and on the maximum I/O rate test, a 512-byte transfer request. On both tests we set the read/write distribution to 100 percent read and the random/sequential distribution to 100 percent sequential.
We performed our custom test in three ways. The first included a 512-byte transfer request size and distributions set at 100 percent read and 100 percent sequential. The second was a 2-KB transfer request with a 67 percent read-33 percent write distribution and 100 percent random. The third was a 64-KB request set back to 100 percent read and 100 percent sequential.
NetIQ Chariot Test
NetIQ Chariot's Ethernet test employed Cisco Systems' 3550 12-port copper Gigabit Ethernet switch, Extreme Networks' Summit7i 24-port copper Gigabit Ethernet Switch, two Cisco 7200 VXR routers and two Cisco 2948G-L3 10/100 switches. The storage network included a Qlogic SANbox-8 Fibre Channel switch, nStor Technologies' NexStor 18F Fibre Channel array with 18 18-GB Fibre Channel Seagate drives; a Qlogic QLA-2300 Fibre Channel host adapter and three Maxtor 5,400-rpm ATA/100 hard disks.
In addition to the devices under test, this test included an HP Compaq DL 580 server and a Dell Precision Workstation 410.
Caw Networks Web Avalanche
We used Caw Networks Web Avalanche 4.0 to measure transactions per second under Microsoft IIS, untuned. We set up this test with a 10-second ramp-up time, 100 Caw users per second. File sizes were set to 7 to 10 KB. Other settings include http 1.1, MTU 1500, and two retries on retransmit.
Our test bed for this suite included two Dell PowerEdge 2500s with 512 MB of memory, 36 GB of disk storage, and two 600-MHz Pentium III processors; 12 Dell Optiplex GX1 workstations with 256 MB of memory, 9 GB of disk storage and one 600-MHz Intel Pentium III processor; and eight generic 1U workstations with 256 MB of memory, 10 GB of disk storage and one 600-MHz Intel Celeron Processor.