This chart from Pingdom shows how downtime at Tumblr has mounted in recent months, compared to other blog platforms.
Although some fast-gowing web services (like Netflix) choose the cloud computing model, most of the time they lease additional data center space. A case in point: the insta-blogging service Tumblr, which has had a dismal uptime track record lately, includng a 24-hour outage. Tumblr said Friday that has raised $30 million in additional funding, and would use some of the cash to expand its data center infrastructure.
"After the holidays, we’ll be opening our second data center to nearly triple capacity and prep for much faster growth next year," writes Tumblr's David Karp. "Today, we already light up more than a rack of new servers every week."
Tumblr is the latest social media company to encounter reliability problems related to hyper-growth. The service is now serving more than 500 million pageviews a month. But its infrastructure has been unable to keep pace, as documented last week by Pingdom in an analysis of uptime at major hosted blog hosting services.
"Tumblr has had a rough time with their stability lately, which is made abundantly clear by this survey," Pingdom wrote, noting uptime of just 96.8 percent for hosted blogs. "Some Tumblr blogs were down for a total of more than two days during the two-month period of this survey. The Tumblr blogs we monitored had an average of more than 300 outages during these two months, some very brief, indicating an ongoing performance issue with the service. Tumblr also had the single longest outage by far, where its blogs were unavailable for almost 24 hours on December 5-6."
The Dec. 5 incident was tied to the failure of a database cluster. The site has also had to cope with denial of service attacks.
"We feel for Tumblr, because it’s quite likely that they’ve had issues scaling the service in face of its growing popularity," Pingdom noted. "This rapid growth is bound to be a factor in the downtime we’ve seen."
The new funding gives Tumblr the resources to add infrastructure and hire five new engineers. It will be interesting to track whether its uptime improves.
