A SuperNode Oops…

24 hours later and everyone is still talking about the Skype outage from yesterday.  And while everything is almost back to normal we should look at how important Skype has become to consumers and enterprise users. Skype recently announced that they had hit the 25M simultaneous user mark worldwide.

As of a little while ago Skype has about 16.5M users back online.  Enterprises must be squirming right now, and Skype is learning how hard it is to be an essential part of the business world.  While we are not perfect on AIM, making sure that we don’t lose the entire network is really important.  Part of the reason that AIM, and other networks like Microsoft, Yahoo and Google can handle and overcome outages or downtime is due in part to our architectures.  Having a centralized network hosted in multiple datacenters around the world allows us to quickly migrate users if we lose part of the network due to equipment failure.

In Skype’s case their own architecture was their undoing.  Skype has a system that is distributed via a series of nodes.  Machines that are in more friendly environments act as SuperNodes where Skype clients connect.  According to Skype “a handful of Windows clients failed and set off a chain reaction that brought down Skype.”  A full post mortem on the outage still needs to be done, but its clear that if Skype wants to work with enterprises it may need to rethink the backbone that powers the service.

Here is a great link describing the Skype architecture.

Here are some more stats GigaOm compiled this afternoon on Skype:

Here is the video from Skype CEO Tony Bates updating everyone on the outage:

2 thoughts on “A SuperNode Oops…

  1. Pingback: Google goes all in on WebM | Sum of the Web

  2. Pingback: Why Microsoft Had an $8.5B Over-Reaction with Skype | Sum of the Web

Leave a Reply

Your email address will not be published.