IT is really a pretty hard job, like it or not, we seem to take abuses from all directions: management, irate customers on the phone, idiot users etc. But you shouldn't have to put up with additional frustration from the tools you use. The sad part is, many of you are fighting with the management tools you use daily, but don't often realize it because you're used to it, and possibly because you have low low expectations from management tools in general.
What am I talking about? Case in point:
- Patient #1. The "I Love My Packet Sniffer" Man I know I've ragged on Packet Sniffers in previous posts, but this one actually came from a true story. We were onsite with a prospect, asking, probing, talking about their network and
application management challenges. Apparently they had a fairly severe problem the week before, but it was "ok" because our "Packet Sniffer" man said "yeah, I was able to track it down in 4 hours with my packet sniffer". Great, customers had severe issues for 4 hours, and that's not a problem at all? - Verdict: he's fighting with his tools and didn't even know it.
Patient #2. The "I Have An Agent for That" Man This is not one person really, but a composite of a few guys we had talked to in the last couple of month. And yes, the name is indeed inspired by the Apple ad/meme. For those of you that are not shy about throwing agents around to monitor everything, consider this: Let's ignore for the moment that agents are a pain to install, maintain, upgrade, just think about the effect on the system. The problem is, agents perturb the very system they are meant to monitor, and the effect is cumulative. So you throw on an agent for an app-monitoring tool, and another one for a different tool, before you know it, there are 5 or 6 agents stack on top of each other, skewing any results you get back from monitoring. - So yes, in this case, you're fighting with your agents and didn't even know it.
Patient #3. The "Behold the Synthetic Transaction" Man I used to work for a company that pioneered the use of synthetic transactions in the testing space and successfully extended that to monitoring. So I definitely believe in synthetic transactions, and know that they have a place in the performance management universe. However, synthetic transactions inherently have an under-sampling problem and simply can't deal with intermittent issues. By the basic definition of intermittent errors, you're more likely than not to miss them if you only use synthetic transactions for monitoring. Not to mention that real world applications will often fail in ways that differ from the "synthetic failures", so purely relying on them is like driving with a huge blind spot and not bothering to crank your neck to look at the real traffic situation.
So this is what we think: if you spend more than 10 minutes tracking down a problem with your existing tools, or if you don't have all the visibility you need as the application moves in real time, you just might be settling for a comfortable old tool that may not cut it any more. Be open to alternatives, and don't be stuck fighting with your management tools and not even know it.