High availability network components

The foundation for a highly available environment is dependent on the network that is created for this environment. The WebSphere Application Server high availability manager, by default, uses a channel framework network protocol model to establish network connections. This model enables network
ports to be shared among all of the channels within a transport chain. A unicast protocol, which supports communication between a specific receiver and a specific sender, or a multicast protocol, which supports open communication to all receivers, can also be used. The following components are important parts of a high availability network:

Distributed Consistency Services (DCS). This component implements JVM level group services. It is responsible for detecting when JVMs in a core group start and stop. It also tracks the set of servers that have joined the core group. The set of servers currently in the core group is a called the view. A JVM typically joins the view in 20-30 seconds after it initializes. This delay allows multiple JVMs to start together and to join the view together rather as a sequence of new views, each with a single additional member. High availability resources can only be activated or deactivated in a JVM after the JVM has joined the view.
The high availability issues message HMGR218I to the system out logs whenever a view changes. This message occurs simultaneously across all current view members, and indicates the number of current view members. A view changes whenever a JVM stops or fails, and whenever a server fails or leaves the view. A view change enables the high availability manager to quickly react to failures and recover any services running on the failed server.
Reliable Multicast Messaging (RMM). This component is a very high-speed publish and subscribe message transport that DCS and the high availability manager use for JVM to JVM communication. It can operate using any of the following low-level transports:
- Reliable Multicast. This is true multicast.
- TCP unicast. This emulates multicast over unicasts.
- Channel Framework. This emulates multicast over channel framework point to point streams.

A high availability environment ties all of the network components together. It provides the basis for providing peer-to-peer failover. The moment a component loses its network connection to the rest of the infrastructure, the high availability manager assumes a failure has occurred and assumes the work assigned to that component.

Thousands of high availability groups can exist within a core group. The policies that are associated with a high availability group control how and when members of that group are activated. A group services mechanism is used to communicate high availability group membership information between members of a core group. This information includes the group services that are available and the lightweight component groups that are part of a high availability group. The group services mechanism also supports the following types of messaging between core group members.

High speed First In First Out (FIFO) messaging between members of the lightweight component groups.
High speed view synchronous messaging.
Java Virtual Machine (JVM) heartbeats or pulses, which are used to indicate that the associated application server is still healthy.

Searchable topic ID: crun_ha_netcomp