Skype: Client Updates May Have Played a Role in Service Outage

The Skype representative who first invoked the phrase "perfect storm" to describe the confluence of events that led to 48-hour service outage on August 16, told BetaNews today that circumstances such as those we experienced in our own tests yesterday could indeed have contributed to triggering that storm, if only in a minor way.

Villu Arak, who works in Skype's global public relations department, told BetaNews this morning that the incident we experienced, and which we described to Skype at length, was the first he'd heard about unusual client-side update behavior. "Still, while what you described could have played a small role in the disruption," he wrote us, "the series of events that led to the outage was triggered by unattended automatic reboots."

On Monday, August 13, Skype sent out a general e-mail to its customers - including me - that a new version of its software was available for download. I didn't download it that week...actually, I waited until I had some time to devote to it, and until the "perfect storm" had already passed. That day came last Monday, when I installed the new version on one of my systems. For the record, it actually worked fine, and continues to do so.

But two days later, from out of nowhere, the system popped up a message asking me to reboot the system to complete Skype installation. If the behavior is the same for other users, it could explain the strange time gap between either the Monday notice or Microsoft's Patch Tuesday (which Skype believes triggered the storm), and Thursday when the service outage occurred.

Arak has reason to doubt it, though. "'Human-assisted' reboots that you described could not have happened with such a high level of synchronicity," he wrote. He later added a reminder that the morning rush hour in Europe where Skype is based corresponds to 3:00 am Eastern Time in the US, "3am being the default time when Windows updates occur to booted machines."

We then followed up with another question for Arak: Is there any physical evidence from Skype's tests which would distinguish an "unattended automatic reboot" from a "human-assisted" reboot? Or in short, how do they know for sure without looking at the system logs of client systems...logs which, last we checked, Skype doesn't have access to?

Arak's response politely indicated to us that this question had not been asked before, and no one really has the answer right now. "But let me assure you that our engineers have chased every lead and turned over every stone," he added.

Despite the lack of available new information, Arak's initial response to us is the first indicator we've heard from Skype that the company is willing to consider the possibility that a small part of its client software - not even the communicator itself, but frankly just the installer package - could have contributed to its service outage last week.