3 Comments

Summary:

Streamy launched in 2007 and officially closed in 2010, but the social news reader application is back in a stripped-down version — not to make its founders boatloads of money on its own, but to prove the capabilities of their new Hadoop application platform.

For a short window between March 2009 and January 2010, the world had access to a social reader service called Streamy that promised to improve on the standard news reader via the personalization power of the social graph. The company and the service closed shop shortly thereafter (it was founded in 2007) and its founders went their separate ways (Jonathan Gray to Facebook and Don Mosites to Zynga), but now they’re back — kind of — with a new service called Streamy Lite.

However, although Streamy Lite is a fully functional social reader service (more details on that below), it’s not a direct money-making venture this time around. Rather, Gray and Mosites are out to prove the power of their new startup, big data platform-as-a-service Continuuity (Gray is co-founder and CEO, and Mosites is a UX designer). The first time around, building Streamy on top of HBase turned out to be a chore that took up far too much time (Gray explained some of this in a blog post), but Continuuity’s platform, called Reactor, made it a mere side project, Gray said.

All told, building Streamy Lite — which is up and running on a five-node Hadoop cluster — took just two weeks compared with months the first time. Managing Streamy Lite takes just a couple hours per week but Streamy required a full-time system administrator. Nitin Motgi, Gray’s co-founder at Continuuity and one of the engineers responsible for building Yahoo’s front page recommendation system, said Streamy Lite is almost equal to the first version of that system in terms of capabilities and infrastructure, but the work at Yahoo took about six or seven months.

streamy2

A platform inspired by Facebook

As a social-graph-powered service, Streamy probably owes its entire existence to Facebook — and Streamy Lite certainly does. As Gray explained, Facebook — impressed by the work he did abstracting the Streamy application from the guts of HBase — hired him after Streamy closed to build out its messaging platform atop HBase. It was the work he did there making HBase easier to manage that inspired Continuuity.

“When we did this at Facebook, that’s when I first realized what this company was going to become,” Gray said.

Jonathan Gray Continuuity Structure Data 2013

Jonathan Gray at Structure Data 2013. (c) Albert Chau itsmebert.com

Continuuity, as we’ve explained before, is akin to a platform-as-a-service or an application server for big data. The various Hadoop data stores (such as HDFS and HBase) and complex programming are hidden from users via Reactor’s high-level APIs and graphical interface. This is valuable because many big data applications really require the same sorts of functionality when it comes to management.

For example, Gray said, “there’s a tremendous amount of commonality” between what he built at Streamy, at Facebook and now at Continuuity. Because even an advanced big data user like Facebook hasn’t rolled out a company-wide development platform (“If that [specific] team didn’t built the tech, they’ll want to build their own,” Gray said), he and his co-founders (both of whom came from Yahoo) spotted an opportunity to help developers everywhere focus on the application rather than the infrastructure.

“Using Hadoop is not going to be your differentiator when everyone is using Hadoop,” Gray said.

A reader that shows how the sausage is made

As for Streamy Lite, the service, it’s intended to be a “viable application” and provide a “compelling” user experience, but is by no means intended to fill the gap left by the demise of Google Reader, Gray said. It currently uses Facebook Connect to power its social aspect, but Continuuity plans to keep improving the service and its capabilities over time.

Streamy Lite is also intended to show off what’s possible using Continuuity’s platform. The service consists of separate components for crawling the web and ingesting news content; analyzing the text of the content; targeting and ranking the content; and analyzing readership data. Users experience this via features like exposed data around clicks and clickthrough rates for each article in a user’s feed and for Streamy Lite overall, and the topics and behavioral history that shape their personal profiles. All the topics and named entities within the platform are indexed and searchable, as well.

streamy1

Developers who want to try their hand at improving on Streamy Lite can clone the code right from the service and have at it.

“We don’t want to make it seem complex, but we do want to show why is this interesting, why is this cool,” Gray said.

Yes, it’s ultimately a Continuuity PR stunt; but at least it’s a good one. That Streamy Lite was built and is up and running is validation of the company’s approach providing a big data development platform. On the off chance it — or even someone else’s forked version — takes off, we’ll see how well Continuuity’s platform can handle the type of scale that comes along with a popular web application the likes of which Continuuity presumably hopes to host for its customers some day.

Check out the video below to see the original Streamy.

  1. hmm…interesting. our team here at opentray.com will have to give it a spin, considering our business is aggregating Big Data news.

    Share
  2. Stefan Groschupf Tuesday, October 22, 2013

    Checkout Jon at Big Data Brews talking about the Streamy History:

    Share
  3. “lite” ?

    Share

Comments have been disabled for this post