Stack Exchange's Really Big Monitor Setup(blog.stackoverflow.com) |
Stack Exchange's Really Big Monitor Setup(blog.stackoverflow.com) |
1.) Even though the person who hired us fell in love with the set of NORAD from the movie Wargames with walls of giant monitors, not one of the NOC analysts sitting at their desks ever once looked up at the big screen (unless it was late at night and they replaced the network map with a first-person shooter) to do their job. They always look at the event viewer at their desk.
The big wall of screens is purely to impress customers/investors during visits.
(Though we do want one but only because they're cool.)
I'd love to hear some war stories and see how you work with other teams.
If you want to say that this is a managerial dashboard wankfest that may indicate a company more obsessed with metrics than with building a great product, just say it.
Edit: whoops, didn't notice it was $$$$.
I know either is a small fraction of the cost of the overall device, but you'd still want some kind of benefit for the price increase.
http://www.newegg.com/Product/Product.aspx?Item=N82E16814121...
These guys are hooking in to their analytics package. The web analytics services we connect to include Google Analytics, Chartbeat, GoSquared & Mixpanel natively, but you can use our push & pull APIs to connect custom widgets.
I've been out of the network management space for about a decade, but ostensibly the big screen was pitched as a way to look at overall network health or identify a large-scale outage. There'd be a big-ass network topology map that would suddenly start showing red dots.
For the big topology map, that was usually done using Spectrum or Openview Network Node Manager (the only use we had for NNM for that matter, it sucked at everything else).
The reason I said "ostensibly" though, is because if you're relying on noticing red dots appear on the map to tell there's an outage, it means you aren't doing proper event correlation.
Most systems we built focused on the event viewer (Netcool for the most part, but it was just for deduplication as it didn't do actual correlation); so that's what would generally be on screen for an analyst. Their other screen would have their ticketing system (usually Remedy, but occasionally Vantive, which I vastly preferred).
The actual correlation engine was Nervecenter for the most part (which was a consultants dream as it took like 6 months to build all the rules), although we tried out a bunch of the "code book"-based correlation engines (SMARTS I believe was the name of one of the more egregious ones).
I worked on one project to build a NOC for Raytheon that 6 months into it they decided to make the whole thing automated (as in, not have any analysts actually working in the NOC) and that was interesting. We ended up having to write logic to handle all the notifications and queuing.
Like I said though, all of this was the Cretacious period for network management, so I really hope things have gotten better in the interim.