On October 6, 2004, IBM announced the latest release of WebSphere, version 6. The next day, Jack Martin, editor-in-chief of WebSphere Journal, sat down to talk with Dr. Bob Sutor, the director of WebSphere Foundation Software, about some of the new features in this release.
Websphere Journal: Bob, tell us about the latest news on WebSphere:
Bob Sutor: We're launching WebSphere Application Server version 6, the first major release in nearly two years, and it contains, really, a tremendous number of new features that we think our customers will really be excited about.
WJ: What's your favorite feature?
Bob: The new high-availability manager. You know, if the customer's really critical business applications are running on an application server, it can cost a tremendous amount of money if that server somehow becomes unavailable.
For example, there are really two areas that you're concerned about when servers become unavailable. First, if they're not there you can't give them new work. You need to very quickly shift that work elsewhere. The second area of concern is what do you do with those jobs that were going on when the server suddenly became unavailable?
Say you're doing a big international money exchange; is that caught in limbo? How long does it take you to get that up and running again? We've been able to reduce the downtime for some transactions from what might be 5 or 6 minutes to perhaps 10 seconds or less.
WJ: How are you doing that?
Bob: It's a number of things. It's new algorithms; it's the configuration of the WebSphere clusters; and it's making sure that they are configured with network addressable storage. It's having this new manager actually being part of WebSphere. In many cases, this gets rid of the need for external high-availability managers. Since it is right in WebSphere, it has a more innate understanding of what the app server is doing and so can manage and monitor many of the internal mechanisms.
WJ: Can you describe what this manager is and how it works?
Bob: First of all, it works with a number of servers. This is definitely a feature of the WebSphere Application Server Network Deployment product. This is an extension of the failover and workload management features we had before. What the manager is doing initially is monitoring the health of the different servers, noting if they become unresponsive in any of several different ways. When it determines that one or more servers - it could be an entire WebSphere cluster - is no longer available, it can shift the work that was intended for those servers somewhere else in that data center or perhaps to another data center far away. Then what it does is it goes in and it grabs the transaction log, assuming it is available in shared storage, and hands it off for completion of in-flight work. So the new features are about greater flexibility that yield better availability, and these translate to real business benefits.
Bob Sutor - Dorector of WebSphere Foundation Software
