Linux Today: Linux News On Internet Time.

D.H. Brown Study Reveals Tier 1 Linux Cluster-software Failover Products

Dec 11, 2001, 10:15 (0 Talkback[s])

Press Release -- PORT CHESTER, N.Y.--Dec. 10, 2001-- D.H. Brown Associates, Inc. (DHBA) has released a report today detailing the results of a Linux cluster-software failover product evaluation.

Six of the 43 products evaluated are stand-outs: Mission Critical Linux NetGuard 0.1.1 beta, GoAhead Self Reliant 9000 2.0 beta, Vertel eORB 2.1 and Eternal etEngine 1.2b2, Legato Cluster Enterprise 4.7, Steeleye LifeKeeper 3.01 beta, and PolyServe LocalCluster Enterprise 2.1. The study found that these offerings accomplish Rapid IP failover in times ranging from 0.04 to 3.8 seconds.

Companies need to recover capabilities and information quickly when disasters strike. Failover requirements are stricter than ever before. Linux-based clustering is growing in consideration and use in telecommunications, banking and finance, e-commerce, manufacturing, and other industry segments where high-availability systems are needed. Linux cluster-software failover time is one of the key variables of interest among the numerous products offered to accommodate Linux-based clustering.

IP failover times provide a baseline for an actual application-failover latency. (There are of course, many other test points, features, and functions to consider.) Application-failover times depend on the amount of state information to be restored, as well as many other factors. If the IP failover time is, for example, less than one second, then it may be fast enough to be insignificant for many applications. However, if the application failover itself must be less than one second, the lower the IP failover latency, the more time remains to failover the application itself within its constraints.

The report also details the requirements of the key blocks of a proposed future hardware/software stack for such clusters. This hardware stack includes such critical elements as application management, cluster protocol, fault-management service, cluster load-balancing service, message service, cluster event-notification service, and virtual-interface architecture (VIA). All of the stand-out products provide for remote administration - a critical feature in many Linux deployments - and each offers cascading failover, another vital capability. Data concerning significant features and functions is provided for all these products and dozens of other Linux cluster products.

Link to the page where the report may be purchased and the executive summary PDF