If you can't measure it....

If you can't measure a quantity that you need to examine for your job or current task, then you're focusing in the wrong place.

When it comes to computing, visibility is key to maintainability, serviceability, reliability and success. Visibility means measurement.

This is a concept not lost on those who have come before us, given the multitude of packages and protocols dedicated to this end.

I'm lucky enough to work with people who understand the critical importance of measurability and monitoring. Unfortunately, the industry isn't lucky enough to have had this most basic and fundamental fact become a pervasive truth. Let's take a simple example.

Today, I reinstalled Microsoft's latest attempt at a desktop OS, Vista. My desktop rig has 2 RAID controllers, both of which have WHQL accredited drivers available.

Anyway, the initial leg of the install went perfectly. Upon rebooting and Vista actually launching from the hard disc, I was greeted with the box rebooting at a certain point. While I figured out what the problem was and how to fix it, would an error message of any type or description really be too much to ask? Even a stack trace is more helpful than a reboot.

At a higher level, I have taken care of many networks where monitoring and measurements amounts to testing ping response times and little else. And the examples of this happening were for very big, very valuable networks. Networks of over 100 devices have not had even this level of monitoring.

Thankfully, the excuses of mundane, involved setup routines for monitoring are all but gone these days. Take Zenoss and Hyperic - both excellent packages that provide excellent visibility and monitoring of many different types of software and both easily customisable and extensible.

If you have a system that involves two or more interconnected programs, hosts, networks, switches, routers, firewalls or any other type of device may I suggest you monitor it - for your own sanity if nothing else?

Correlation is best seen when nice uniform measurement is in place and "just happens". The number of problems you find may not change, but the time to resolution and the preemptive options you have will increase and there are no exceptions to this as long as your monitoring is extensive and complete.