[AIX Solaris HP-UX Linux Windows]

Considerations for large topologies

WebSphere® Application Server can run optimally in large topologies. Considerations in the use of the HA manager, core groups, and Intelligent Management are important for ensuring optimum performance for large topologies.

HA manager
Several components of the HA manager require configuring and must be tuned appropriately for large topologies.
  • HA manager protocols
  • HA manager memory buffers
  • HA (active) coordinator

To learn more about important HA manager configuring and tuning, see Top 10 things to know about High Availability Manager (HAManager) in WebSphere Application Server.

Core groups
  • Single core group topologies require configuring and tuning the HA manager.
  • Partitioned cell multi-core group topologies require configuring and tuning core groups.
  • Intra-cell bridged topologies require configuring and tuning the intra-cell bridge.
  • Intra-cell bridged topologies require configuring and tuning the intra-cell bridge.
Core group scalability is improved by setting the following custom properties: Setting these properties enables WebSphere Application Server protocol implementations to improve scalability and administrative usability not only in environments larger than 50 members, but also in smaller environments.
Consider this summary of recommendations for core groups:
  • Avoid one large core group (DefaultCoreGroup).
  • Partition the cell into several smaller core groups. Keep core groups in the 50 JVM size range.
  • Turn off the HA services for one core group that holds JVMs that do not need the HA manager.
  • Turn off ODC if the proxy server, Web services clients, remote request dispatching, WebSphere Extended Deployment, or ODC routing are not in use.
  • Bridge between core groups rather than increasing the size on one core group.
  • Dedicate one JVM in a core group to be coordinator and bridge (not the node agent).
  • To overcome the limitation of having an administrative JVM in each HA manager-enabled core group, create “phantom” nodes to host additional node agents that can be added to new core groups.
  • Select the appropriate core group protocol version.

WebSphere Application Server cells larger than 50 WebSphere JVMs, application servers, node agents, deployment managers, or proxy servers must consider important factors in the planning and deploying process.

Intelligent Management

A key component of the Intelligent Management is the on-demand router (ODR). The ODR is a proxy server based on Java that proxies both the HTTP and SIP protocols. SIP is not supported on the z/OS operating system. The ODR supports health, application edition, and performance management features. It can manage both WebSphere and non-WebSphere environments. The ODR can queue requests for less important applications so that requests from more important applications are handled quickly.

For WebSphere Application Server - Network Deployment, the following Intelligent Management functions are enabled by default.
  • On demand router (ODR), routing rules and policies
  • Dynamic load-balancing
  • Autonomic controllers, which provide dynamic cluster, dynamic workload balancing, overload protection, and elasticity support
  • Health management
  • Application edition manager
  • Dynamic clustering
  • Intelligent management enabled web servers
  • Service polices, which provide traffic prioritization

These Intelligent Management functions require more memory and CPU usage, depending on the number of JVMs and nodes in the cell.

If you do not want to use the Intelligent Management function, or do not want to incur any increased memory requirements or CPU utilization, you can set the LargeTopologyOptimization custom property to false. Setting this custom property reduces the overall cell memory and CPU utilization. For more information see, Cell custom properties.

To learn more about Intelligent Management and the benefits it can provide in your environment and operations see the following: