25 Things I Hate About Your Network25 Things I Hate About Your Network

One of our favorite network troubleshooting gurus, Terry Slattery, has put together a very cool-looking network diagram showing the 25 Biggest Network Problems. Not surprisingly, virtually all of them are either directly or indirectly relevant to real-time/voice traffic. I talked with Terry about some of the high- (or low-) lights.

One of our favorite network troubleshooting gurus, Terry Slattery, has put together a very cool-looking network diagram showing the 25 Biggest Network Problems. Not surprisingly, virtually all of them are either directly or indirectly relevant to real-time/voice traffic. I talked with Terry about some of the high- (or low-) lights.The 25 problems aren't listed in any order of frequency or severity, and Terry came up with the list based on his consultations with Netcordia customers. One of the big categories of the items on the list relates to configuration issues, which Terry says are endemic. He quoted a colleague's statement that, "The major cause of failures are fingers."

His best advice for avoiding configuration-related and other human errors: Peer review--Did you interpret the condition correctly? Is this the right time and place to install the change? "They can't just be cowboys," wanting to ride in and save the day and ride out again, Terry said: "If you pull the trigger too quickly, you could end up with a smoking hole in your foot."

To call out just one of the other 25 problems--I hadn't heard a lot about Duplex Mismatch recently, but it's on Terry's list as a problem he still sees often. This is where one end of a connection is set to full-duplex and the other half-duplex, resulting in packet loss that Terry said can reach 30%.

Terry said that duplex mismatches persist in large part because different vendors have different defaults. "It's a nagging problem," Terry said. "In my estimation, the [auto-negotiation] protocol is broken." By that he means that the protocol was written when 10-Mbps half-duplex was the common implementation; now it's 100-Mbps full. Under the current protocol, when there's a duplex mismatch, the connection defaults to half-duplex. Terry's suggestion is that the protocol be updated so that if the auto-discovery process finds it's a 10-Mbps connection, the connection would go to half-duplex, on the assumption that it's an older network link. On a 100-Mbps connection, it would default to full duplex.

If you want to check out the poster with the complete list, it'll cost you your email address; the download is at Netcordia's website. Here's a slice of what it looks like: