Updated Redundancy for MassStore
This is a whiteboard discussion of updated redundancy for MassStore.
In this short video I’ll describe how Masstech’s updated architecture focuses on redundant processes rather than machines, upgrade options for existing non-redundant systems, and redeployment options for sites presently using redundancy.
First, a quick review. MassStore is composed four core processes: The app server, the database, the datamover and HQS transcode. In MassStore’s simplest deployment, all four are installed on the same physical or virtual server. And of course these are all attached to storage.
The new minimum redundant configuration runs these same four core processes on two server platforms. Again, either physical or virtual.
The app processes are clustered. The other processes run in an active/active configuration. If one of the redundant processes fails, regardless of which server it is on, the other continues in single-thread mode and operation is not interrupted. When a process returns the system automatically switches back to redundant operation.
A bonus of this configuration over stand-alone is that when operating in redundant mode (i.e. no failures) an extra datamover and transcode engine are both available.
Customers with a one-server non-redundant system can now achieve full redundancy by simply adding a second server.
Let’s compare this to the older minimum redundant configuration which required four servers.
This configuration was focused on machine redundancy rather than process redundancy. Though constantly running and ready to take up load, no process on the redundant server was available until a failure was detected in the primary. When that happened all services switched from the primary server to the secondary. The database required two additional servers for redundant operation, though neither of these could host additional processes.
Our focus on redundant processes rather than redundant hardware allows us to consolidate a completely redundant system on only two servers.
Existing customers using the older redundancy scheme may choose to redeploy their systems, either to reduce the number of server resources required, or to increase processing capacity and speed by installing additional datamover and transcode licenses on their extra servers.
Essentially, an older system can be converted into a basic two-server redundant system with at least two extra datamovers and transcoders.