Say you have a database problem that is choking your application. Everybody feels it, and you get plenty of complaints. When you can track it to an obvious source like a database issue, it’s not that big a deal. You get it sorted out in most cases and the application is back up and running smoothly again in a reasonable period of time.
But the toughest problems are not so easy to solve. It’s not a general problem with the pain spread out across the user population. It’s more erratic. Some users complain and some don’t. When it’s sporadic, isolating the issue becomes a nightmare for monitoring professionals because unlike the database issue referenced above, the problem isn’t so obvious and when the source of an issue is unclear, it’s easy to deteriorate into a blame game.
You’re on the phone talking to people and everyone is blaming someone else. It must be users who simply don’t understand or bad programming or a database malfunction or something. (See Top Five Monitoring Excuses, for more information).
That’s why it’s so important that your monitoring software enables you to get to the transaction level to track the exact nature of the issue affecting your web site or applications. The problem with most monitoring tools, no matter what they do, eventually, they show you bunch of averages. The problem with averages, however, is that when you have sporadic problems, the real issues are more often than not, hidden in the averages and fail to point you to the real nature of the issue.
That’s why you need to use monitoring software that looks not only at averages, but lets you look at behavior of transactions across an element. You need to have a concept of transaction behavior modeling that lets you focus clearly on the metric in question and overcome the problem where averages mask the real issue.
At the same time, you don’t want to get caught up at the other end of the spectrum where you are looking at every transaction and you get lost in too much data. So you need to be able to drill down to solve the problems and get at the information that’s most important.
It’s a fine line to walk, but you have to find monitoring tools that let you get at the source of tough problems without getting lost in too much data or stuck with too little.