Terminology is the most important thing to know when when starting out with WebSphere Application Server

Over the last few weeks I’ve done a fair amount of consulting on IBM Connections – not so much the install and technical stuff but more simply talking to customers about WebSphere Application Server (WAS) and how it works. The single thing that people new to WAS seems to struggle the most with is the terminology and getting the overall architecture in place. Once that’s done most people actually like the platform and find it nice to work with. A while back I linked to a PDF containing a nice graphics on slide 4 (Overview of IBM WebSphere Application Server Concepts for IBM Lotus Connections Administrators) but it’s hard to find so I’m reproducing it below.

The first thing and single most important piece to understand about WAS is that a “server” is usually not what you think! 🙂 Or at least not what it meant to be being namely a physical thing. With the advent of virtual machines the WAS definition is getting easier to understand I think. Below is an attempt to explain the WebSphere terminology to someone starting to deal with IBM Connections.

I find that it makes the most sense to start from the physical/operating system layer (Windows server / Linux server). Each machine/operating system instance with an IP address or hostname is a “node“. The list of nodes in your WebSphere environment may be seen in the ISC under “System Administration/Nodes”. On a node there may be one or multiple servers. A server is a Java Virtual Machine (JVM) process meaning that it may be controlled individually and it may have memory assigned to it and have specific JVM parameters set if required. Please note that having multiple servers on a single node is very common in WebSphere Application Server.

When a server is created in WebSphere Application Server it is created using a profile which basically is a template for the server. There are two main profiles to worry about – “default” which is an server to run applications and “dmgr” which is the deployment manager. In an architecture with multiple nodes the deployment manager is the administration server that controls the nodes it knows about including making sure that the configuration is synchronized between the nodes. The nodes belonging to a single deployment manager belongs to the same cell. In other words there is a single deployment manager per cell.

Looking back at the nodes a node may be managed or unmanaged. A managed node is a WebSphere node that the deployment manager can talk to and send commands. An unmanaged node is the opposite. WebSphere knows about the node but cannot communicate with it on a WebSphere protocol. Something like an IBM HTTP Server (IHS) is therefore an unmanaged node. If a node is managed it will have a nodeagent installed. The nodeagent is a separate Java process which is started and stopped individually and separate from any server. The nodeagent can be used to synchronize the configuration to the node and to start and stop “stuff” on the node. The nodeagent may be configured to start all the servers running on it when ever it starts. On servers the administrator may install applications which run on the server. The applications may be started and stopped individually but usually all applications start and stop with the server.

For high availability you may choose to create a cluster of servers. Servers in a cluster runs the same applications and can take over from one another. Note that you cluster servers and not nodes so you could have a cluster of three servers on a single node.


If you have an IBM Connections installation where everything is installed on the same Linux or Windows box and you chose the “small deployment” which puts all applications into the same server you will have the following:

  • A single cell
  • A single deployment manager controlling your cell
  • Three (3) nodes. One for the deployment manager, one for the server running the IBM Connections applications (managed node) and one for the IBM HTTP Server (IHS) (unmanaged node).
  • A single nodeagent
  • A single server running all the applications
  • A single cluster with the server
  • Three (3) JVM processes running – one for the deployment manager, one for the nodeagent and one for the server