• United States

Cherchez l’application

Feb 24, 20033 mins
Data Center

One of the questions I get regularly is, “How can I improve my network capacity planning and forecasting abilities?” Here are some thoughts to get started.

One of the questions I get regularly is, “How can I improve my network capacity planning and forecasting abilities?”

Here are some thoughts to get started:

First, make sure you’ve got solid technical processes for measuring network performance and bandwidth consumption. Specifically, you should look at end-to-end latency across the network. By end to end I mean system to system – not just from the WAN interfaces on the routers. There are a host of fancy tools to do this. If you can’t afford them, at least write a short script to run pings or traceroutes on a regular basis.

You also should be tracking bandwidth utilization. Look at average and peak utilizations, and track both over time. For example, your average utilization might be growing at 8% month over month, while your peak utilization grows at 15%. (Hint: You’ll probably need a network upgrade sooner than expected). Also look at average utilization as a percentage of peak utilization, and note whether that figure changes over time.

You should look at the sampling rate at least hourly (every 15 minutes is better), and you should review historical trending at least quarterly.

OK, let’s say you’re doing all that. Guess what? That’s only half the story. Knowing what’s going on is table stakes for playing the game. But to win, you need to understand why it’s happening. Do this in two steps.

Step 1: Make sure you understand which network resources are being consumed by which applications and how that changes over time. Is HTTP growing faster than e-mail, for instance?

Step 2: Time to find out about macroevents that are driving your organization’s use of the network. For example, is your organization engaged in data center consolidation? If so, you’ll probably need to think about increasing the bandwidth to the remaining data centers – and providing redundancy and reliability alternatives. Time to research your local metropolitan-area providers.

What about a Web services or IP telephony rollout? You’ll need to check out latency requirements, and you might want to invest in compression technology (particularly at remote branch offices).

How will you know about this? That’s the tricky part. You’ll need to leave your team to their scopes and Sniffers, and begin asking questions. Start with your boss, but don’t stop there. Talk with executives in other departments (sales, accounting and customer service).

And don’t limit your conversations to the senior folks. Midlevel staffers often have a lot of great information – and more time to talk. That clerk in accounting might provide a valuable heads-up about the financial package slated for rollout next year. The visiting salesman might clue you into the new office the company’s considering in Latin America. And so on.

Of course, you’ll need to validate this information before acting on it, but as they say, knowledge is power. Knowing that these options are potentially in the works can help you make better decisions today.