It is no secret that service-oriented architectures are complex with wide ranging service interactions via messaging
middleware. This week open source Linux vendor RedHat and traditional software management vendor CA, announced new products to provide insight into complex services interaction and allow better performance tuning.
Also this week, IBM released a new version of WebSphere Application Server, with runtime provisioning capabilities that allow more services to be deployed on fewer processors.
JBoss, a division of Red Hat, announced today that it is melding its JBoss Enterprise SOA Platform and JBoss Operations Network to provide remote monitoring and management of open source SOA deployments.
On Tuesday, CA announced enhancements to its data center automation management products including its Application Performance Management (APM) technology. The CA Wily APM tools are now optimized to manage complex and highly distributed SOA and virtualized environments for both .NET and Java since CA is platform neutral.
John Rymer, analyst with Forrester Research recently noted that all the major vendors including IBM, Microsoft and Oracle are also developing tools to deal with SOA complexity.
"The fundamental issue is complexity," Rymer said. "It's so complex to build and manage these SOA applications."
The shift to SOA from client/server applications of the 1990s and the simpler Web apps of the early 2000s, brought up issues that IT departments had not dealt with before, said Prabhjot Singh vice president of APM technologies for CA.
"As soon as you make a service available to the world you really have no control over how it is going to perform because it may be called a hundred times more than you intended," Singh said. "It isn't as easy anymore because it might be hosted on another continent by a third-party vendor so the complexity has grown."
Burr Sutter, senior product manager for SOA at JBoss, explained the complexity of processes within services.
"A service is made up of component parts," he said. "In a service, the initial step might be transformation. Take the inbound data, transform it to data that we need down stream. The next step might be to call the rules engine to process logic against this data. That might call an EJB component or do some routing, for example, send this message to two other endpoints, or send it to a third party service endpoint."
System administrators need to monitor the services and messaging at that level to find bottlenecks.
"Each step has its own performance characteristics and own message throughput," Sutter explained. "We track it down to that level. So you can see if the transformation component is the slow one in the mix. Or the routing or the EJB might be the slow down."
The JBoss monitoring tools check the processing time for each of those components and the message load, including how many messages succeeded through that channel and how many messages failed," he explained. There is alerting for both performance bottlenecks and messaging failure, he said.
The systems administrator can look at performance characteristics of processes and messaging over a period of time, such as a month, and the JBoss tools automatically provide a baseline, Sutter said. Then an alert can be set up if the number or messages goes down by a certain percentage, which may indicate a problem.
In the 24/7 mode that most SOA applications operate in, the systems administrator can receive alerts even at home and access the console from a PC or notebook via a corporate VPN to view and remedy the problem, Sutter said.
For its part, IBM has described a new rev of Websphere Application Server that allows easier remote monitoring of distributed servers. The version of Websphere Application Server, said Marisa Shumway, program director, Websphere Application Server, IBM, has runtime provisioning that "dynamically selects only the needed functions for memory and space."
The result, she said, is dramatically reduced by the application server foot print.
Central in the new set up is a Network Deployment manager which lets IT control "large-scale distributed topologies."
"You can look at them and manage them from a single location," she said.
With reporting by Jack Vaughan