Cloud Computing Economics - There Is No Free Service

Cloudonomics Journal

Subscribe to Cloudonomics Journal: eMailAlertsEmail Alerts newslettersWeekly Newsletters
Get Cloudonomics Journal: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn

Cloudonomics Authors: Lori MacVittie, Skytap Blog, David H Deans, Shelly Palmer, Tim Crawford

Related Topics: Cloud Computing, Web 2.0 Magazine, Cloudonomics Journal, Facebook on Ulitzer, Cloud Data Analytics


Facebook Rates Tilera ‘Swell’

Facebook has been testing Tilera’s last-generation widgetry and rated it the industry’s best performance per watt

Guess Tilera can stop all the double-talk about whether or not Facebook - and the other major-league cloud companies - except the not-invented-here-enamored Google - are interested its newfangled proprietary many-core x86-compatible servers.

Facebook has just publicly admitted it's been testing Tilera's last-generation widgetry and rated it the industry's best performance per watt - better than Intel at common data center apps and at running the fabled Memcached so important to Facebook et al.

The social networking giant said Monday that Tilera servers are 4x better at performance per watt than classic four-core x86 Nehalem servers, leading to the logical conclusion that Facebook will start deploying Tilera servers in production once Tilera's 64-bit next-generation machines are available.

Such an intuition is hardly rocket science. The things are supposed to be 10x better than legacy x86 architectures. What would you do?

According to the Facebook study, Tilera's current low-power many-core processor family delivers a 3.1x advantage over Xeon systems and a 4.2x advantage over Opteron systems in performance per watt when running key value-store applications like Memcached, which is also important to folks like Twitter, Zynga and Flickr.

The testing was done using a S2Q Tilera-based server built by Quanta (which makes Facebook's own Open Compute platform).

The server incorporated eight of Tilera's second-generation TILEPro64 processor in a 2U.

The Tilera server and the competing x86 servers both met Facebook's latency requirement of processing transactions in under a millisecond. Both platforms had comparable latencies, but Tilera's processors won because they processed more transactions a second at the required latency.

Facebook found that low-power many-core processors are well suited to Kvstore workloads (read Memcached-like apps) with large amounts of data. It also found that TILEPro64 can yield at least 67% higher throughput than low-power x86 servers at comparable latency. When taking power and node integration into account as well, a TILEPro64-based S2Q server with eight processors handles at least three times as many transactions a second per Watt as the x86-based servers. Even within a single x86 socket and with just four cores, performance scales poorly and can't take advantage of additional threads whereas the TILEPro64 implementation can easily take advantage of (and actually requires) more cores for higher performance. Previous work has mostly concentrated on mapping KV stores to low-core-count so-called wimpy nodes meaning this trade-off implies higher costs for hardware, system administration and fault management of very large clusters.

Facebook reportedly plans to run the same study on Tilera's new 64-bit Gx3000 series announced in June with sampling about now this month.

More Stories By Maureen O'Gara

Maureen O'Gara the most read technology reporter for the past 20 years, is the Cloud Computing and Virtualization News Desk editor of SYS-CON Media. She is the publisher of famous "Billygrams" and the editor-in-chief of "Client/Server News" for more than a decade. One of the most respected technology reporters in the business, Maureen can be reached by email at maureen(at) or paperboy(at), and by phone at 516 759-7025. Twitter: @MaureenOGara

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.