Things just aren't as simple as they used to be. There was a time when we could achieve nearly all of our uptime goals just by clustering our backend database servers and maybe a handful of other application servers. There was very little of that weird virtualization voodoo going on, and all our important
applications ran on physical servers.
Well, times have changed. Not only are a larger share of our applications considered mission critical, but our service level agreement expectations are much higher and we're under pressure to virtualize more applications in order to realize greater cost savings in the data center. To further complicate things, those applications have grown more complex with more moving parts. For those of us concerned with high availability and disaster recovery, this has presented some interesting challenges.
Consider, for example, an online retail application. That "application" is really a business service consisting of several applications forming a multi-tier application stack. The stack will typically include a web tier, an application tier and a database tier. Now, put that database tier on a big iron Solaris server, the application tier on Linux servers and the web tier on virtualized Windows servers and you can see the problem with traditional clustering products.
Now, here's where I get to talk about my favorite part of the Veritas Storage Foundation/High Availability 6.0 release due out in early December.
Concurrent with the 6.0 release is the release of Veritas Operations Manager 4.1 and Veritas Application HA 6.0. VOM is the centralized management interface for Veritas Storage Foundation and Veritas Cluster Server and for a couple years now has provided operational visibility into all VCS clusters running in a given environment, across all data centers and OS platforms. Symantec introduced Application HA a little over a year ago to provide application-level HA for VMware environments. In this release, VOM introduces support for a construct called Virtual Business Services, and AppHA is extended to cover KVM, Solaris Oracle Virtual Machines (formerly known as Logical Domains, or LDOMs) and IBM AIX Logical Partitions (LPARs).
The Virtual Business Service (VBS) construct allows Symantec's customers to link applications running under VCS control (known as service groups in VCS terms) with service groups running in other VCS clusters as well as virtualized applications running under AppHA control. With VBS, I can define the dependency relationship between the constituent tiers of of an application stack, as well as the desired restart behavior when a tier fails over in response to a fault. The beauty of this is that it makes no difference which OS platform or which virtualization technology an application (or application component) is running on. The VBS concept is supported with VCS Global Clusters to properly orchestrate wide-area failover of a VBS across data centers in response to a site-wide failure.
Finally, in keeping with the promise of building the private cloud you want from the infrastructure you already have, customers at releases of VCS 5.1 or later don't have to upgrade to VCS 6.0 to take advantage of VBS. All that's necessary is VOM 4.1 and deployment of the VBS packages to each VCS cluster and AppHA virtual machine that will participate in a VBS.
Watch this space for future posts (and maybe a demo video or two) on Virtual Business Services. I think you'll like what you see.