BTM (Business Transaction Management) is starting to appear on the radar screen of many clients and vendors. BTM is based on the ability to trace a transaction path through n-tiers of infrastructure components in order to provide 1) visibility into the transaction, 2) a template that could be used to understand how the infrastructure supports the transaction and 3) a basis to define whether a transaction behavior in normal, that is within the resource usage bracket observed historically or abnormal, signaling a performance or availability issue. Many products have appeared on the market over time to support this transaction tracing or transaction tracking ability. Bristol Technology (now HP), MQ Software Q!Nami (now BMC) supported MQ Series and Websphere MQ. ClearApp (now Oracle) and dynaTrace have this capability built in their SOA-APM monitoring solution, Optier and Correlsense provide visibility into transaction paths. Now we find this capability as one of the requirements for Application Performance Management: IBM ITCAM has looked into this issue from the get go, CA is developing the capability for Wily, HP is working on expanding Bristol, Compuware and Opnet can take advantage of their network analysis solutions to provide this information, Quest Software has added this feature in their APM solution and Precise is also walking the same path . Why? Simply because the focus of APM has shifted from the management of critical code (as in Java EE) to more complex n-Tier composite applications and that clients are pushing for an end to end, holistic view of the applications that goes beyond Java EE or .NET monitoring. This is even more critical when it comes to virtualization: different metrics have to be collected in virtual machines and it becomes more critical to know where the transaction is in a such a fluid environment. Another point is the benefits that such solutions would bring to a virtually consolidated data center: while a lot of the so called "data center automation" focuses on the provisioning and moving of workloads, the fundamental reason for allocating resources, e.g. the "trigger" that tells the administrator whyresources should be differently allocated does not exist yet, and this is precisely where BTM could provide precious information. This is a fundamental element of data center automation, and beyond it process automation, that has been neglected.
What strikes me as very important here is the fact that this is precisely what we expected of BSM in 2005 and 2006. I can recall discussion with major vendors such as BMC, Mercury and IBM about the fact that ADD (Application Dependency Discovery) and the resulting CMDBs were not stand alone solutions but should be used to improve system management and sold with it. In my paper "BSM 2.0" I tried to analyze the reasons for BSM to have moved from its original trajectory to become what it is today.
On the other hand, BTM seems to have taken over and is actually tracking the original BSM route. BTMis not BSM 2.0, but at least it is part of it, and a very pragmatic part of it that should attract many clients. The only thing missing (but I know it is on the agenda of several vendors) is the capability fro transaction tracing to feed dependency data into the CMDB. Of course, to achieve this the CMDB should first become a CMS as discussed by my friend and colleague Glenn O'Donnell.