Access and location of virtual volume data

Best performance is achieved by accessing virtual volumes data resident in the local cache.

Access of virtual volume data typically adheres to one of the following scenarios:
  • If a recall is required to place the virtual volume in the local cache, it is done as part of the mount operation.
  • If a copy of the virtual volume is not available at a local TS7700 Cluster, and a copy is available at a remote TS7700 Cluster in the grid, then the volume is accessed through the remote TS7700's cache. A virtual volume is not available on a cluster if that cluster does not have a copy or if the copy it does have is inaccessible due to an error.
  • If the local copy requires a recall, and a remote copy is already resident in cache, remote access can occur if the remote access penalty is measured to be lower than the time to recall the volume locally.
  • If a recall is required to place the virtual volume in the remote cache, it is done as part of the mount operation. If access is through a remote cache, the available data rate depends on the available bandwidth provided for the inter-cluster links and the distance between the TS7700 Clusters.

Grid network load balancing

For a TS7700 Grid link, the dynamic load balancing function calculates and stores the following information:
  • Instantaneous throughput
  • Number of bytes queued to transfer
  • Total number of jobs queued on both links
  • Whether deferred copy throttling is enabled on the remote node
  • Whether a new job will be throttled (is deferred or immediate)
As a new task starts, a link selection algorithm uses the stored information to identify the link that will most quickly complete the data transfer. The dynamic load balancing function also uses the instantaneous throughput information to identify degraded link performance.