Pandaboard Cloud Cluster Running Google App Engine

Noritsuna Imamura showed an ARM cloud cluster built with 6 Pandaboards at  Linaro Connect Q4.11. Noritsuna is a member of the Open Embedded Software Foundation (OESF), a Japanese organization that support Open Source embedded software.

Google App Engine (GAE) on ARM Server (12 Cores)
Cluster of 6 Pandaboards Running GAE

The cluster is made of 6 panda boards with a total 6 GB of RAM (1GB per board) that runs the Google App Engine (http://code.google.com/appengine/) in Ubuntu/Linaro 11.09 release. The middleware used is TyphoonAE, a full-featured and productive serving environment to run Google App Engine (Python) applications. This type of server hosts web applications such as Rietveld (used in the demo below) to be run on (thin) clients. There are also plenty of other software (required by Google App Engine) that runs in this ARM cloud such as MySQL, Apache2, memcached and more.

In the video, a power meter shows that the cluster consumes about 35 W in low activity mode.

Norisuna had a comparison table that showed the Pandaboard cloud (OMAP4 @ 1GHz – 12 Cores) uses 4x times space and 5x less power (50W vs. 257 W) than a Xeon 1.8GHz Quad Core server with about the same amount of processing power and memory (6GB vs. 8GB).

One drawback is that the software is located on the SD Cards of the pandaboards so it can be relatively slow, but this could probably be replaced by SSDs connected via USB without affecting power consumption too much.


The presentation slides are available at http://www.slideshare.net/noritsuna/panda-cloud

Share this:
FacebookTwitterHacker NewsSlashdotRedditLinkedInPinterestFlipboardMeWeLineEmailShare

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK 5 ITX RK3588 mini-ITX motherboard

11 Replies to “Pandaboard Cloud Cluster Running Google App Engine”

  1. @cnxsoft
    Yes the purpose of this farm is to build the Ubuntu distro, that take a long time on ARM proc, even if they get more power, more cores 😉

    I’m a ARM addicted, and I LOVE your blog 🙂

    Regards
    Vincent

  2. “…Xeon 1.8GHz Quad Core server with about the same amount of processing power and memory…”

    any data to back up that claim? Is that just raw MHz or real benchmarks?

    1. That’s a big assumption from me, no benchmarks here:
      Xeon: 4 cores at 1.8 Ghz = 7.2 GHz
      6 OMAP4: 12 Cores at 1 Ghz = 12 GHz

      I could be wrong here, maybe I should not have written that…

  3. @av500
    I went to Coremark website and compared the Coremark 1.0 scores of OMAP4430 running 2 forks and Intel Xeon E5405 (Quad Core @ 2 GHz) running 4 threads.
    OMAP 4 gets 5742.20 and Xeon E5405 gets 23579. (See screenshot below)

    http://www.cnx-software.com/wp-content/uploads/2011/11/benchmark_coremark_1.0_xeon_e5405_vs_TI_OMAP4430SDP.jpg

    But I still think the Xeon server would probably perform (much) better than the pandaboard cluster during “real” tests because of the storage used in the pandaboards (SD Card or external USB storage).

  4. @av500
    It really depend on what you need …

    – For pure performance -> Xeon
    – For scalability, efficiency, no power consumming -> ARM

    Using ARM teach me 1 thing : EFFICIENCY ! For example, my blog is self-hosted on an ARMv5 DockStar with USB DD and 128Mb of RAM, I have to optimize image size and compression level, using caching, no database, small php cms engine … To deliver contant even faster than some other web server using Xeon 🙂

    On my blog I’ve made some benchmarks, I can serve 2000 unique visitors / 12h !!! And actually I have only about 500 uv/day -> http://www.it-wars.com/article243/performance-de-l-auto-hebergement-avec-le-dockstar My blog post is in French but use the GoogleTranslet widget in the sidebar and feel free to leave comments 😉

    Regards,
    Vincent

Leave a Reply

Your email address will not be published. Required fields are marked *

Khadas VIM4 SBC
Khadas VIM4 SBC