Business Transaction Management Blog

I read a discussion last week on the internet where a person asked a simple question “How can I get an End to End latency breakdown of my business transactions”. Before I’d even had chance to comment I noticed that two software vendors had already replied with links to their website claiming that they deliver exactly what the original poster was asking for (imagine the odds of that eh? ).

I’m not one to start a fight but I’ll gladly point out a few things which might help customers understand what several vendors mean when they claim “End to End” coverage of monitoring business transactions.   Firstly,  they’ll simplify things by saying “URL to SQL”, they’ll then tell you they can provide this visibility by just sticking an agent on each of your J2EE and .NET application servers. With just two tier agents you’ll magically get your “End to End” latency breakdown and all the visibility you need to solve all of your problems and a whole lot more. In fact I know a few vendors who will instrument your coffee machine if you ask them nicely (go the extra mile and all that).

To the Java or .NET developer this type of proposition sounds very credible as they get  ”End to End” visibility of exactly what they are interested in as transactions pass in to and out of their JVM’s and CLR’s. The problem is that business transactions don’t just execute through J2EE and .NET tiers. Ah I hear you say but my APM vendor can show me latency breakdown to databases, message buses and outbound web service calls. Yes, that is true but all they are doing is timing the outbound protocols of things like JDBC and SOAP from the JVM or ADO.NET/Remoting from the CLR – its not exactly rocket science to time an API call from a JVM or CLR these days. But what happens if your business transactions continue to execute in tiers beyond JVM’s and CLR’s? What happens when one JVM publishes a message to an ESB and then several other applications subscribe? What happens if you get other business transactions that invoke tiers without going through a JVM or CLR? A batch job is a business transaction, a marketing report is a business transaction – how do you track these entities when all you can see is your JVM and CLR activity???

You can only see what you can see. If your monitoring tools only focus on J2EE or .NET then guess what you see? You only see your information from a J2EE and .NET context. Business transactions and your applications span more tiers than the average person thinks. I’ve witnessed customers draw visio diagrams of what they think their application environment looks like. I’ve witnessed the same customers in complete awe once they’ve seen what diagrams a BTM solution will draw. Customers see tiers they never thought were integral parts to their applications, they get a true “End to End” view of how their business transactions traverse their infrastructure.

Here is a quick topology diagram of what “end to end” views customers can expect to see from a BTM solution. Diagrams like these are automatically generated from business transaction flow and contain tiers beyond J2EE and .NET (and their outbound tier calls). Seeing the forest instead of few tree’s helps customers really understand the latency breakdown of their business transactions. It also helps them understand the real IT dependencies that their business services have.