9.4.1 One server in the cell is a member of one bus
In this topology, there is only one bus. There might be multiple appservers in the cell, but only one is a member of the bus. This is roughly equivalent to the typical V5.x JMS server topology.
The pros of this topology are:
- It is very simple to set up and manage.
- It can be expanded later by adding more servers to the bus.
The cons are:
- Clients running on other appservers in the cell have to connect remotely to the bus rather than connecting locally. This can affect messaging performance.
- Clients running outside appservers have to connect to the bus member to do messaging. The connection factory you use needs to have provider endpoints configured with the details of the bus member server.
If the SIB service is enabled on other appservers in the cell, then connection factories can be configured with provider endpoints that point to a list of bootstrap servers. See 8.7, Connecting to a service integration bus for more information about using a bootstrap server and defining a list of provider endpoints.
In either case, all messaging connections go to the bus member server and might affect messaging performance.
- Message consumers might not be on the same server as the queue points they are consuming from. This could have a performance impact.
- This topology cannot be upgraded to support high availability or workload management. High availability and workload management require clustering appservers. You can create a new cluster and include the bus member as the first appserver in the cluster. However, this will not automatically give you the messaging high availability features that are normally associated with adding a cluster as a bus member.
- Using the bus member server as the template for a cluster server is not equivalent to adding a cluster to the bus. No bus information is copied as part of the template process. The SIB service will be enabled on the new cluster server as a server property, not part of any particular bus.
- Using the bus member as the first server in cluster server is not equivalent to adding a cluster to the bus. Only the original server is part of the bus.
It is possible to add a cluster to the bus, delete all of the queues you want to be highly available or workload-managed, and recreate queues of the same name that have their queue points located on the new cluster bus member. Any messages on the queues are lost when they are deleted.