Image from Thomas LaRock
I had a great time over the last month writing a series of posts with my friend John Herbert (@MrTugs) over on the SolarWinds Geek Speak Blog. You can find the first post here. John and I explored the idea that people are always blaming the network for a variety of issues that are often completely unrelated to the actual operation of the network. It was fun writing some narrative prose for once, and the feedback we got was actually pretty awesome. But I wanted to take some time to explain the rationale behind my madness. Why is it that we are always blaming the network?!?
Visibility Is Vital
Think about all the times you’ve been working on an application and things start slowing down. What’s the first thing you think of? If it’s a standalone app, it’s probably some kind of processing lag or memory issues. But if that app connects to any other thing, whether it be a local network or a remote network via the Internet, the first culprit is the connection between systems.
It’s not a large logical leap to make. We have to start by assuming the the people that made the application knew what they were doing. If hundreds of other people aren’t having this problem, it must not be with the application, right? We’ve already started eliminating the application as the source of the issues even before we start figuring out what went wrong.
People will blame the most visible part of the system for issues. If that’s a standalone system sealed off from the rest of the world, it obviously must be the application. However, we don’t usually build these kinds of walled-off systems any longer. Almost every application in existence today requires a network connection of some kind. Whether it’s to get updates or to interact with other data or people, the application needs to talk to someone or maybe even everyone.
I’ve talked before about the need to make the network more of a “utility”. Part of the reason for this is that it lowers the visibility of the network to the rest of the IT organization. Lower visibility means fewer issues being incorrectly blamed on the network. It also means that the network is going to be doing more to move packets and less to fix broken application issues.
Blame What You Know
If your network isn’t stable to begin with, it will soon become the source of all issues in IT even if the network has nothing to do with the app. That’s because people tend to identify problem sources based on their own experience. If you are unsure of that, work on a consumer system helpdesk sometime and try and keep track of the number of calls that you get that were caused by “viruses” even if there’s no indication that this is a virus related issue. It’s staggering.
The same thing happens in networking and other enterprise IT. People only start troubleshooting problems from areas of known expertise. This usually breaks down by people shouting out solutions like, “I saw this once so it must be that! I mean, the symptoms aren’t similar and the output is totally different, but it must be that because I know how to fix it!”
People get uncomfortable when they are faced with troubleshooting something unknown to them. That’s why they fall back on familiar things. And if they constantly hear how the network is the source of all issues, guess what the first thing to get blamed is going to be?
Network admins and engineers have to fight a constant battle to disprove the network as the source of issues. And for every win they get it can come crashing down when the network is actually the problem. Validating the fears of the users is the fastest way to be seen as the issue every time.
Mean Time To Innocence
As John and I wrote the pieces for SolarWinds, what we wanted to show is that a variety of issues can look like network things up front but have vastly different causes behind the scenes. What I felt was very important for the piece was the distinguish the the main character, Amanda, go beyond the infamous Mean Time To Innocence (MTTI) metric. In networking, we all too often find ourselves going so far as to prove that it’s not the network and then leave it there. As soon as we’re innocent, it’s done.
Cross-function teams and other DevOps organizations don’t believe in that kind of boundary. Lines between silos blur or are totally absent. That means that instead of giving up once you prove it’s not your problem, you need to work toward fixing what’s at issue. Fix the problem, not the blame. If you concentrate on fixing the problems, it will soon become noticeable that networking team may not always be the problem. Even if the network is at fault, the network team will work to fix it and any other issues that you see.
I love the feedback that John and I have gotten so far on the series we wrote. Some said it feels like a situation they’ve been in before. Others have said that they applaud the way things were handled. I know that the narrative allows us to bypass some of the unsavory things that often happen, like argument and political posturing inside an organization to make other department heads look bad when a problem strikes. But what we really wanted to show is that the network is usually the first to get blamed and the last to keep its innocence in a situation like this.
We wanted to show that it’s not always the network. And the best way for you to prove that in your own organization is to make sure the network isn’t just innocent, but helpful in solving as many problems as possible.