In a recent study by Aberdeen Group, four out of five network managers said they’d spent hard cash on network management tools in the last two years. The same study revealed that only two out of five felt they now had a better understanding of their network as a result. The report’s conclusion was simple : it’s easy to get more performance data from your network – but getting useful information is way harder.
It’s because of difficulties like this that testing and measurement of enterprise networks has, if not a bad press, a poor reputation amongst those who have responsibility for their smooth running. It’s too easy to drown in the volume of data such tools produce, and so they’re often left in a corner and only brought out when there’s a fire to fight.
This is completely the wrong mindset. Testing and Measuring your network should be something that’s going on all the time, establishing a baseline ‘feel’ for how things are during normal operation. Only then will you be able to spot differences when there’s a problem: a new application appears, or a relocated group of users suddenly reshape your traffic patterns. These may sound obvious, but more recent research proves that most enterprises simply don’t know the basics about their network: what applications are running over it, who’s using it, or what sort of performance they’re getting. Think about your network for a moment, and see if you could give a 60-second briefing, right now, to your management on the above points.
The reason so many organisations get away with this is simple. Like users, Network Applications have an annoying habit of just getting on with things. Engineered to cope with unpredictable performance, they don’t break, but just gradually get slower and slower. Over a long period, users get accustomed to poor performance: no-one benchmarked it, so no-one spots the deterioration - and there’s always more pressing tasks to be thinking about.
When things do finally grind to a halt, the situation is likely to be really bad. Asking a management team to invest in new technologies like WAN Acceleration, or MPLS, from this position of ignorance is a sure recipe for wasted time and money. How can you fix a problem when you don’t really know what the problem is? In this environment, new technology becomes the equivalent of a weight-loss video or self-improvement book - it looks like you’re doing something about the problem, but actually all you’re doing is spending money.
In defence of network managers, having to deploy monitoring and measurement tools is doubly irritating when you consider that most businesses would like to be thinking about their networks less and less, not more and more. They’d really just like to get on with running their business, and have the network as a simple, reliable utility like power or water - a viewpoint Service Providers are keen to encourage.
Viewed in this light, though, simple monitoring makes more sense. Most people keep a high-level picture of their utility usage such as power consumption (increasingly so, thanks to ‘green’ initiatives) or the number of phone lines they use and the average time to answer an incoming call. It makes sense to start thinking about your network in the same way, and define some very simple metrics that you can benchmark against. Start to view Test and Measurement as a continuous, but most importantly, a simple activity that doesn’t drown you in data. You’ll get to know your network in peacetime - and you’ll be much better equipped to catch those inevitable fires when they’re small.
This article originally appeared in Network Computing magazine, June 2009
