The world doesn't revolve around IT. Consultants like yours truly sometimes have a hard time grasping that concept, because our lives really do revolve around IT. So when I walk into a site where we helped build the network three years ago to find that the total amount of network monitoring that's been going on since we finished UAT (user acceptance testing) is ... nada, you can imagine the initial reaction.
But they're customers, so my throwing things around wouldn't be appreciated. Instead, in response to complaints about network slowdowns, I simply ask a few questions concerning bandwidth numbers on specific WAN links, latency numbers from the switches, or even just CPU utilization numbers off the servers.
When I don't get an answer, I can dig up that big green documentation notebook I left behind during the UAT (after blowing the dust off it, of course) and point to that section on Things To Do Going Forward. That little part in there on deploying desktop, network and server monitoring tools.
This one client of mine didn't completely ignore that advice, but what it deployed was Ipswitch's WhatUp Gold -- a real nice package, but one that's mainly famous for basic up/down end-node device monitoring. The software actually can do some bandwidth and network health monitoring, but when we get called in for emergency work, we like to use what we know, not eat the learning curve on something new.
For us, that used to be MRTG (Multi Router Traffic Grapher), but lately we've turned to Cacti (http://cacti.net/index.php). A big reason is it's free. A second reason is it's pretty quick to set up. It also delivers its information via color graphs that are easy to show to clients for proper effect. Writing down "100% server CPU utilization" for example, doesn't have nearly the impact of showing a graph that jumps to 100 from 15 with a big red pointer. Nifty.
Cacti gets this advantage from a technology called Round Robin Database (RRD). This system stores and displays time-sensitive data in Cacti's MySQL database. The information expands over time, but Cacti stores it in such a way as to keep storage requirements extremely low. The result is those sexy graphs with very little overhead required to run the app.
Another advantage for which we like Cacti is that it's much easier to configure than MRTG. First, the app comes with a load of preconfigured generic device templates that'll cover most of your bases out of the box. Then, configuration is managed via a friendly Web-based GUI, not the command-line syntax that often accompanies MRTG. Now for professional network geeks, the command line isn't a disadvantage. But it becomes a disadvantage when you hand it off to your customers. Either they eat the learning curve, or they flood your phones with configuration, modification, and interpretive requests. Cacti is easy enough that training is a one-day affair, and after that, the system can be handed off entirely.
Sure, they may eventually move to something more enterprise-oriented, such as SMS and MOM (Microsoft Operations Manager), but for companies that need a quick fix and introduction into why network and server monitoring data is so useful, Cacti is a great first step. After that, it's extensible via new device additions, scripting and even SNMP, so it can carry even medium-size businesses for quite some time.
Yeah, this is the Windows Enterprise column and I'm gushing about a Linux-based tool, but in my defense, Cacti does run on Windows. You just need to compile it appropriately and use Cygwin for part of the configuration. I think. To be honest, we've never set it up on anything other than Fedora Core. Hey, what can I say -- the Penguin has power.