“The network is slow.”
Probably the worst complaint a Systems Administration team at any small to mid-sized office can get. The end users often can’t pinpoint what “slow” is or when it happens, it’s seemingly random, or they report it after the fact when there is nothing to actively troubleshoot.
I am not a networking “guru” by stretch of the imagination. Like many small offices, our NetOps team consists of several people who may have some areas they enjoy or “specialize” in, but are mostly jack-of-all trades, ready to jump in and sort things out whenever things need attention. I enjoy the variety, but sometimes the ongoing project list leaves you in a situation where certain areas of your “kingdom” are left until they cry out in pain.
The LAN in my office was one of those lost souls. Sure, I’ve got my Network+ training, I used to have a valid CCNA certification, I know the difference between a hub and a switch and I can find enough of the settings in my HP and Cisco switches to assign IP addresses for management access and use some basic features. And then my skill set drops off there – because small networks are often “set it and forget it”.
We think about collecting SNMP logs and monitoring traffic and all that cool stuff and then reality sets in: I wish I had the time to spend installing and learning enough about those tools so they can be really useful when someone comes knocking with a “slowness” complaint. But I don’t. So finally I brought in someone who actually looks at networks every day. Someone who knows the settings on network gear and can look at how they work together. Yes, I can pull out some crossover cables and make packets move from point A to point B, but I wanted some advice from someone who really understood how it all worked.
It was eye-opening. My switches that linked the users workstations to our servers were all connected, but they were naturally oversubscribed without taking advantage of trunking any of the ports together to pass traffic to core switch over larger pipe. Spanning tree was configured incorrectly and not at turned on all on some switches.
The end result was that while my Layer 3 setup looked fine to me, the Layer 2 traffic was actually taking an extra hop through a switch that was accidentally acting as the spanning tree root, adding unnecessary delay. After correcting that issue and ordering up some gig modules to add trunking up to our core switch, upload/download speeds of files to servers appears to be coming close the maximum available from the desktops.
Next up – increasing the speed of our internet connect by switching from frame relay to fiber from our ISP and subscribing to a bigger pipe on that end.