1. Administration Guide
  2. Implement Session Delivery Manager on Your Network
  3. Cluster Data Distribution Techniques

Cluster Data Distribution Techniques

An OCSDM cluster groups server nodes to offer reliable access to OCSDM system application services without disruption when failure condition(s) occur. The cluster support ensures that all submitted tasks can be processed as long as one member node is available.

The following data distribution techniques are used on an OCSDM server cluster to configure devices, deliver messaging information, protect and maintain the cluster database, and transfer files within the cluster:

  • Device configuration—Any cluster node can go to a device on-demand to retrieve its most recent configuration. This allows each node to be synchronized to have the same configuration version, which prevents the need to replicate large datasets between nodes.
  • Message driven data—Some data sets that can be subject to network latency such as fault management events, polling statistics and audit trails are distributed through a Message-Oriented Middleware (MOM) service, which provides asynchronous processing to an OCSDM system, so that this system can scale resources both vertically and horizontally. The resilience of the MOM is maintained by guaranteed deliveries and durable subscribers. The data is generally stored on the local host machine in a dedicated local database.
  • Database replication—Sensitive data sets such as local configuration view (LCV), user security, and device management are transactional and need to be available over the cluster are maintained by a database replication group, which maintains one master database in the cluster at all times while all other members are replicas. Retrieval of datasets is done on the local host machines. However, transactional modifications to the data are done on the master database, which then replicates the transaction to the replicas. Replication keeps the cluster database synchronized. If the master database fails, the remaining replicas elect a new master database.
  • Push-pull file transfer—Large datasets such as route sets, which are maintained in the local database or file system, are transferred around the cluster through push and pull mechanisms. Host nodes use the MOM service to publish that datasets are available through events, while other nodes use SFTP to pull information from other nodes or push information to other nodes.