Abstract

This paper documents a series of experiments performed on LVS by Mission Critical Linux, Inc. in a cooperative effort with Intel Corporation. The objective of these experiments was to evaluate LVS's ability to distribute web requests among several servers. We investigated a variety of LVS configurations and offer a comparison of LVS's ability to scale on Linux 2.2 versus Linux 2.4. In contrast to similar evaluations, our entire test effort was accomplished using open source software on Linux based platforms.

Our results show that in a uni-processor environment the performance of LVS on Linux 2.4 is on par with Linux 2.2, however in a multi-processor configuration, Linux 2.4 significantly surpasses Linux 2.2. LVS on Linux 2.2 actually exhibits minimal scaling in a multi- processor environment. We reveal the detrimental impact that multiple devices sharing interrupts can have on LVS throughput. A comparison of LVS to a commercial load balancer indicates that LVS is a viable alternative to the more expensive, proprietary solution. Our results show that LVS is nearly twice as cost effective in terms of price/performance when compared to the hardware based load balancer. Lastly, we document the steps necessary to enhance the capabilities of our load generator which in turn reduces the amount of client hardware needed.

• View the full text of this paper in HTML, PDF, and PostScript.
  The Proceedings are published as a collective work, © 2001 by the USENIX Association. All Rights Reserved. Rights to individual papers remain with the author or the author's employer. Permission is granted for the noncommercial reproduction of the complete work for educational or research purposes. USENIX acknowledges all trademarks within this paper.

• If you need the latest Adobe Acrobat Reader, you can download it from Adobe's site.

• To become a USENIX Member, please see our Membership Information.