Filed under: Uncategorized
Uptime is king. We at Port80 agree 100%. If we were professional sports players or coaches, we’d even agree 110% (if that makes any sense at all). However, what exactly is Internet “uptime”, in the perfect sense? Can your server be “up”, and yet your Web site or domain be considered “down”? Of course, any hiccup any ol’ place along the network between you and the observer or client makes you look down. Then consider that people may hit your site from all over the world — add it all up, and that’s a whole lot of potential for instability or perceived instability.
Of course, most site uptime observation (that you pay attention to) is done by your own monitoring system or maybe from fairly stable monitoring points on the Internet at large (if you employ some third party service). Now, even in the later case, you really don’t see what everyone sees: it is better than just pinging yourself, but even having outsourced monitoring doesn’t cover everyone out there who could access your site. You can’t reliably see what some random pocket of some ISP or phone company is doing to its customers in Akron, Ohio because some of their routers went crazy, unless you are in Akron. What about the DNS server that crashes at a local big company? Maybe there’s a misconfigured proxy at the apartment building with the shared T1? Or your “borrowed” neighbor’s wi-fi goes offline mid-request? On and on, the possibilities add up, leading to an old saying in a new context: “Never will there be a day on the Internet that there won’t be some problem some place.” Of course, this is a safe statement, as you could say the same for electrical power. It really does go off (not just when there is a solar storm) and much more often in some parts of the world than you might think.
Now, why all this ranting and raving about uptime and trying to understand from whose perspective we should measure this stuff from? Well, a relatively new social monitoring service for servers, www.pingdom.com, recently released a report about popular site downtime (http://royal.pingdom.com/?p=116) we found quite interesting. As you will see, even the big guys have a few hours of downtime, at least from the point of view of the site’s members’ monitoring. Our question would be to measure what a less distributed environment says about downtime, as well as what the monitored sites themselves have to say about it. We hazard to guess the numbers won’t be quite the same from each network perspective. We also hope that the distributed service isn’t divergent from backbone and local monitors because, if so, the “I’m not down, you are!” concept really does have merit.
All kidding aside, it is clear that end-user perspective matters, but how many users “down” makes a site really “down”? We guess it depends on who and where those users are versus what is happening to those users’ network routes during the failure… or something like that. Let’s hope that Pingdom can pick up a large and wide user base, because this really could help show the stability/instability of the Web to everyone interested. That’d be us.