You can download the PPT source for these diagrams here

Use this template to help create environment diagrams. In my personal experience it is an area performed badly by Middleware Administrators.

This template is my way of providing some help to make creating diagrams of your environments less painful when you are under pressure to perform your support role.

It is in PPT as this will allow a wider audience, not everyone in your organisation has VISIO.

This diagram set is based on IBM WebSphere Application Server Network Deployment and Liberty-Runtime (Core) Collective and only shows a few examples of fictitious Cells. There are many designs that could be drawn, these are just some examples.

Single Cell, (Single Deployment Manager). Contains two nodes, thus two managed-node profiles.

Two application clusters. Each cluster contains two JVMS (servers) for an HA service concept, but all installed on one physical host for testing purposes. This is known as a vertical cluster. A vertical cluster is when one or more of the entire cluster members exist on the same physical machine.

Single Cell, (Single Deployment Manager). Two nodes, thus two node profiles. Two application clusters. Each cluster contains two JVMs and thus presents two copies of each application for service.

This is also known as a horizontal cluster. When one or more members of a cluster exist on separate hosts (via managed nodes) then the cluster is considered to be a horizontal cluster.

Complex Highly Available WAS ND environment. We can can see in the diagram below that we have many managed nodes.

The green coloured managed-nodes could even be considered to be hosted in a secondary facility i.e. second data-centre to ensure high availability. It will all depend on your HA design and ability to route traffic across data centres. Either way, the design allows for multiple host and JVM failure and provides a level of tolerance.

Not that the weakest point is the fact that the Deployment Manager is not highly available, however this is not required for a running HA environment unless maintenance/configuration changes need to be actioned.

A simple way to cater for a Deployment Manager corruptions is to ensure that you have quality consistent backups of the DMGR file-system.

It is possible to perform a Highly Available DMGR design, however it requires thought and is a delicate procedure to get right and you should consult a WebSphere Specialist to help with a highly-available Deployment Manager scenario.

Solution to provide HA of “App” during deployment of a new version. We can stop Cluster Members from communicating (take off-line), and or App’s and release a new version as required. Then using controlled administration ie cluster-weights, we can ensure that testing is able to verify the new application is working before allowing full distrubution(Sunchronisation) across the cluster members.
Solution to provide HA of “App” across a “4 Member Liberty Collective”
Note: 2 Physical hosts are depicted in this example

Libery-Runtime is not the same prodcut as WAS 8.5.5.x ND, and so the pattern for deployment is completely different, as is the automation and configuration using Shell and Jython.

We can leverage some of our WAS 8.x.x.x knowledge, however understanding a new set of API’s and XML configurations will be required.


Leave a Reply