DataSynapse Upgrades Virtual SOA Clustering

DataSynapse today announced version 2.5 of FabricServer, its platform aimed at spreading SOA and similar applications across clusters of servers. DataSynapse describes this as application virtualization, as it's aimed at achieving much the same as standard server virtualization but doesn't have to involve multiple OSs. But in many ways it's closer to grid computing, in that it aims to combine multiple servers into one large computing resource, whereas virtualization is mostly concerned with dividing up a single server.

In DataSynapse's architecture, all servers are initially configured identically, with every required application installed on each alongside the FabricServer software. This combines load balancing with management, the former distributing requests to the appropriate server while the latter allows administrators to divide the total compute capacity between the installed applications. In a large cluster, individual servers would likely be dedicated to a single application, but this can easily be changed as all software is installed on all servers.

Like standard server virtualization, this can make an IT infrastructure more flexible, though unlike virtualization it doesn't by itself help much with consolidation. The disadvantage is that the DataSynapse software needs to be able to support every application it's managing, meaning that it won't necessarily work with everything. The new version adds support for more Microsoft environments, including .Net v3, SharePoint and IIS. It already supported previous versions of .Net on Windows, and most Java platforms running under Linux, both of which can run either on VMware or bare metal.

Most DataSynapse users will see the greatest benefits when using it alongside VMware, as this allows multiple OS images to be installed simultaneously. However, licensing could get expensive for off-the-shelf apps, as each one will need to be licensed for every machine that it's installed on, even if only a relatively small fraction of the available capacity is actually used.

DataSynapse isn't the only company offering this kind of product. The most successful is IBM, though its product is geared toward servers running WebSphere applications, whereas DataSynapse can mix software from multiple vendors. The closest competitor is probably Appistry, whose application fabric also distributes Java and .Net applications across multiple servers but works in a different way, giving developers an API that lets them access the clustering functionality directly.