Redfin surprised a few people Friday afternoon when it filed for an initial public offering, but inside the standard boilerplate of risk factors was something even more surprising: Redfin’s entire site is hosted within a single Seattle data center.
From the filing:
Our website is hosted at a single facility in Seattle, Washington. We do not currently have a back-up web hosting facility in a different geographic area. Should this facility experience outages or downtimes for any reason, including a natural disaster or some other event, such as human error, fire, flood, power loss, telecommunications failure, physical or electronic break-ins, terrorist attacks, acts of war, and similar events, we could suffer a significant interruption of our website and mobile application, which would harm our business. In addition, our website and mobile application could be interrupted even if this facility experiences temporary outages, which could also negatively affect our services and harm our business.
That’s something you don’t see very often in the cloud era, and it would appear Redfin actually chose to move off Amazon Web Services into this single backup-less facility at some point over the last year and a half (see update below). AWS offers Seattle as a location for its Amazon CloudFront content-delivery network, but services like compute and storage are only offered out of Northern California and Oregon on the West Coast.
In January 2016, AWS published a case study involving Redfin, which at the time was running “its entire operation on AWS, using a range of services, including Amazon S3, Amazon DynamoDB, Amazon Redshift, Amazon Kinesis, Amazon Elastic MapReduce, and Amazon EC2 to operate its website, data warehouse, and big-data analytics operations,” AWS wrote in the caption of a video outlining the case study.
(Update 7/7: A source familiar with Redfin’s infrastructure strategy said the working of that caption was a mistake on Amazon Web Services’ part: Redfin is still running all of its business analytics operation on AWS, and it was never running its site on AWS. The caption has been updated on YouTube and on the Redfin case study page.)
A Redfin representative declined to comment on the company’s infrastructure strategy.
There are a few possibilities here. For one thing, risk factor statements in S-1 filings are designed to be written as broadly as possible to protect the company against future lawsuits. Redfin could still be using AWS for data warehousing (a type of database that’s designed for fast reads and analytical applications), which could be backing up its actual data in the event of a disaster, but has shifted the hosting duties to its own facility.
But it’s still fairly surprising that an operation the size of Redfin’s doesn’t have a second facility that would be capable of hosting its site should a disaster (or just an extended power outage) strike the Seattle area. A study from the Disaster Recovery Preparedness Council in 2014 found that “The majority of organizations participating in the survey have made the investment in deploying a secondary site to help manage disaster recovery.”
Once Redfin’s IPO is complete, here’s a suggestion for that new capital: invest in a backup data center.