High-Availability in IT is a strategy to satisfy business availability needs. It is impacted by planned outages and unplanned outages --the result of a system or software fault, which we will not address here. Planned outages are typically a result of a preventive or corrective maintenance task that imposes an interruption to the day-to-day system operation. The traditional approach to planned outage is a careful planning of the intervention to minimize the downtime and the risk of the system not going back up properly. With virtualization, novel approaches can be taken where system downtime and application downtime can be decoupled.
Oracle VM Server for SPARC, f.k.a. Solaris Logical Domains, is a virtualization and partitioning solution supported on Sun CMT servers powered by the UltraSPARC T-class processors. Oracle VM Server for SPARC allows the creation of multiple virtual systems on a single physical system. Each virtual system is called a logical domain (LDom) and runs its own copy of the Solaris operating system. Among its many features, LDoms have the ability to do warm migration between two machines, i.e. to checkpoint and migrate an active LDom from one server to another one.
In ISV Engineering, we have demonstrated this Domain Mobility feature for a running installation of the Oracle 10gR2 database. During the migration, the database server is not shut down. The migration from one physical host to another one is also transparent to the client applications connecting to the database --as long as no timeouts are encountered, or conversely, timeout values can be appropriately set by the application's admin and/or developers for the warm migration to happen transparently-- such that there is no downtime for the application. This work has been documented in the following whitepaper on the Sun Developer Network : Increasing Application Availability Using Oracle VM Server for SPARC.