|
HP ProLiant DL140
Price as tested:$2,733
Warranty: 1 year on parts Web site: www.hp.com |
Engineers evaluated performance by measuring the server efficiency of how it used its memory and processor time while maintaining throughput to network clients. The 10/1000 NIC of the server was connected to a Compaq Gigabit Ethernet switch so that the server NIC simultaneously received HTTP requests from two clients and FTP requests from another two clients.
There were four clients in all, each equipped with a NIC, Microsoft Windows Professional 2000 and Internet Explorer 6.0. Two clients were dedicated to making HTTP content requests, and the other two to making FTP requests. The two FTP clients were each equipped with 256 Mbytes of RAM and instructed to launch 10 requests at a time. Each HTTP client sported 512 Mbytes of RAM and was set to launch 200 threads at a time.
An in-house LoadRun program was used to generate the file requests according to a script written for each client. The order of the file names inside each client script were unique to minimize the possibility that two or more clients requested the same file at the same time, stressing the effectiveness of the server's array rather than cache size.
The second performance test focused on the server's array. The results were obtained using Iometer. The benchmark can read and write data to a storage device according to script. Engineers created an 11-stage script, with each stage having its own settings for data-block size and data-position randomness. As Iometer moved from one stage to the next, the script decreased the block size while increasing the randomness of the data's position.
The first stage simulated a data-streaming application, which typically involves large data sets accessed sequentially. The last stage simulates database and multi-user access, which typically involves excessive seek time and small data sets. The settings of the other stages represent everything between those extremes. All stages were set to produce an equal number of reads and writes.
Each stage was allowed to run for 2 minutes. Data collected during the first 30 seconds of each stage was discarded to eliminate any ramp-up effect.
Discover the benefits of scalable backbone and network services for application and cloud providers.
Network automation is gaining momentum. Here's how you can drive this powerful technology to its full potential.
Subsea cable alternatives can provide a level of comfort for those concerned about disruptions. Realistically, they will continue to play a small, but critical role in the short term.
