X
Tech

The Google Analytics outage: Could there be a reason?

Over the weekend, people had noticed and reported a major Google Analytics outage -- days worth of data aren't available, but Google assures us that our data has not been lost. I wanted to know though if there might be a reason for the outage because it seems strange that it would just up and die when they have so many servers dedicated to processing web stats -- you would think there would be redundancy.
Written by Garett Rogers, Inactive

Over the weekend, people had noticed and reported a major Google Analytics outage -- days worth of data aren't available, but Google assures us that our data has not been lost. I wanted to know though if there might be a reason for the outage because it seems strange that it would just up and die when they have so many servers dedicated to processing web stats -- you would think there would be redundancy.

When visitors use a website that has Google Analytics code, information about the visitor is sent to Google. This mechanism is not broken -- all of these statistics are still being collected and stored in a database somewhere. The problem then is within the servers that crunch that data into compiled units that can be reported on. It does sound like these servers are coming back online and data is starting to show up for some users.

As of 5pm PST this evening (July 30), some users will start to see part or all of the data from the period between Saturday and now appear in reports. We expect updates for all accounts to continue through Monday night into tomorrow and will update this blog when reporting is fully restored.

When I was reading some of the commentary on this latest problem, I came accross this comment on seroundtable.com:

I would almost swear this happened last time right before they had the last PR update. I think they use that data in the algorithm if a webmaster has it. Either that or they defer the resources used to power it.

Is it possible they re-purpose Google Analytics servers to do things like compute pagerank when the algorithm changes? Sounds plausable to me.

Editorial standards