High availability used to require a pair of computers; usually one live and one hot standby. One computer for the price of two. With the limitation of dual ported SCSI disk arrays you could only have two computers able to access a single disk volume (not at the same time). This solution achieved HA but was hardly cost effective. With the advent of SAN and multi-node cluster software, we have been able to design and implement elegant, low maintenance, multiple heterogeneous node, automated HA solutions.
Over the years we’ve had the good fortune to evaluate and use a variety of clustering packages. HP MC/Service Guard, First Watch, Qualix HA+, FullTime Cluster, Legato Enterprise Cluster (now AAM, or Automated Availability Manager), Sun Cluster, and the Veritas Cluster Server. Our product of choice (at the moment) is Legato AAM because of its rich Perl rules based engine, ease of use, complete support for heterogeneous operating systems, enhanced application management functions, and support for any vendor’s underlying disk volume management software (from Sun through Veritas to EMC). With this product we have been able to build “point and click” supportable, multi-node, multi-OS, clusters. The cost of these clusters is not much greater than the cost of similar stand-alone systems. Since you are always using a fair percentage of each computer’s resources there is no additional money spent on spare, idle computers (CPU or Storage). The incremental cost is in the cluster software. By adding AAM clustering and management to your environment you not only gain high availabilty but you gain ease of maintenance. Tier two data center operators and analysts can “point and click” their way through shutdown and startup procedures without needing to consult tier three engineering. By investing in your infrastructure up front you dramatically lower your ongoing total cost of ownership. Strong Link Data has built many clusters with this software, but perhaps the best example is in support of the traditional client/server architecture that many applications utilize. Many IT managers today complain that the cost of the evolution from Mainframes to client/server systems has increased, not decreased, total IT costs. The proliferation of servers and stand-alone islands of storage have driven IT budgets through the roof. With well organized clusters and application availability management you can lower your overhead and regain control of your data center!