As Google Reader winds down, Feedly just keeps growing
Feedly has been adding servers and changing its data architecture — including a move from MySQL to HBase — as more users come aboard and ask the company to start managing many more RSS feeds.
Monday marked the last day of Google Reader, prompting RSS heads hanging on until the end to think about alternatives. While there is a raft of options, Feedly has brought on millions of new users since Google said it would close Reader in March, and it continues to take on more users and information sources while keeping hiccups to a minimum.
Some companies that are already in or are getting into the RSS game — AOL, Digg, LinkedIn — can throw big teams and perhaps their own hardware infrastructure at the challenge of producing a service worthy of taking on users accustomed to the Google experience. But Feedly has been doing this, supporting upward of 12 million users and around 32 million feeds, with fewer than 20 people on staff, said the company’s CEO, Edwin Khodabakchian.
The company has been switching users over from Google’s Reader backend for pulling and serving feeds to Feedly’s own system. Now on that backend, which is called the Feedly Cloud, the system requests the latest content from millions of sources and gets it all ready to send out when users open the application. What’s more, the company has been making available to external developers a Feedly Cloud API, so they can create their own applications with the Feedly-aggregated data.
To do all this, the company relies on Google App Engine for some services, such as processing images to give a quick user experience, but it depends more on its own servers. Since March, among other efforts to support more users, Feedly has doubled its server count to around 100, and the data under management has swelled up to 100 terabytes. The data growth in recent months pushed Feedly to shift from a MySQL and Memcached combination to some Memcached in conjunction with HBase on top of the Hadoop Distributed File System, Khodabakchian said. The new system can support MapReduce jobs in addition to handling real-time workloads, he said.
Those types of engineering changes don’t come for free, so monetization is very much on the agenda. While Feedly plans to roll out information on its monetization plans later this summer, Khodabakchian did talk about how Feedly could serve up information that publishers would want — namely, data on users’ devices, reading patterns and sharing behavior. Right now, it can be hard to know what readers do with it, if they do anything with it. This could happen once Feedly users give their consent for this sort of data to be shared with third parties, Khodabakchian said.
In the meantime, Feedly, based in Palo Alto, Calif., has plenty of other things to work on, such as building out robust personalized searching capabilities and turning out applications for more devices. If things go according to plan, the company could find a nice little revenue stream where Google didn’t.
By Jordan Novet