Abstract

ClickArray Networks, Inc. commissioned The Tolly Group to evaluate its Array 1000, a purpose-built device that combines server load balancing, global server load balancing, reverse proxy caching, SSL acceleration, clustering and Web security capabilities in order to optimize Web traffic performance. The Array 1000 sits between source servers and downstream requesting clients in an Internet or intranet connection. The Tolly Group subjected the Array 1000 to a battery of tests focusing on performance and failover functionality. Engineers tested the ClickArray 1000 version 1.0.1 and determined that the device delivers highly effective Web server offload and hefty performance increases in terms of the number of HTTP requests per second supported.

The ClickArray Array 1000 is a purpose-built Web acceleration appliance designed to reduce load on backend servers by combining reverse-proxy caching, server load balancing, global server load balancing, SSL acceleration, clustering, and Web security into a single platform. In Tolly Group testing, the device was evaluated for HTTP request processing performance and failover behavior in a Web-serving environment built around Windows 2000 Server with IIS 5.0 and WebBench-generated client traffic.  

A primary objective of the test was to validate ClickArray’s claim that the Array 1000 could sustain 20,000 HTTP GET requests per second. Tolly reported that the appliance achieved that level across multiple mixes of HTTP 1.0 and HTTP 1.1 traffic, reaching 20,212 requests per second at a 25/75 HTTP 1.1/1.0 mix, 19,266 at 50/50, 19,537 at 75/25, and 19,969 at 100% HTTP 1.1. Once requested objects were cached in memory, the backend IIS server no longer received repeated hits for that same content, allowing server CPU resources to be preserved for non-cacheable and application-driven requests. The testing used 17 WebBench clients and a single backend IIS 5.0 server, with the Array 1000 configured with 1GB of memory and connected over Gigabit Ethernet.  

The report emphasizes that this architecture can materially improve Web infrastructure efficiency by serving cacheable content directly from the appliance rather than repeatedly invoking the origin server. Tolly characterizes this as highly effective Web server offload, especially for static content-heavy environments.  

High availability was also validated. In failover testing, three Array 1000 appliances were configured in active-standby mode with n+1 redundancy. Engineers simulated failures by disconnecting the outside Ethernet interface on one unit and then a second unit. In both cases, the remaining appliances automatically assumed the virtual IP responsibilities and continued handling sessions without administrator intervention. According to Tolly, when two of the three units failed, the third device successfully picked up the remaining session load, demonstrating resilient failover behavior for Web site continuity.  

ClickArray later rebranded as Array Networks.