1 Comment

Summary:

The furor over Michael Stonebraker’s criticism of Facebook’s scaling of its MySQL database we covered last Thursday has continued to generate comments on the post itself and on Twitter. Friday, Amazon’s CTO Werner Vogels weighed in with a tweet that seemed to accuse Stonebraker of hubris.

Werner Vogels - CTO, Amazon.com - Structure 2011

The furor over Michael Stonebraker’s criticism of Facebook’s scaling of its MySQL database we covered last Thursday has continued to keep that post at the top of our viewer stats and generate comments on the post itself and on Twitter. Friday, Amazon’s CTO Werner Vogels weighed in with a tweet of his own that seemed to accuse Stonebraker of hubris.

Vogels isn’t alone. Other tweets and even blog posts from folks in the industry have accused Stonebraker of throwing Facebook under the bus in order to sell his VoltDB software, such as this one from networking guru and CEO of Boundry Ben Black:

The comments in the post range from those that shed more details on how Facebook uses its database layer to those that show a fairly sophisticated understanding of the tradeoffs of using an existing database technology versus investing in putting a production system on a new and possibly untested new system. However, most of all it illustrates how scaling data has become not just a hot-button issue — but one so complex that it’s difficult to reduce it to cute soundbites like NoSQL or NewSQL. There also seems to be a fundamental schism between those who want the most elegant and appropriate solution and those who want something that works, which they will tweak it as needed. And for technologists, that debate is as old as computers themselves.

  1. Dirk Bartels Tuesday, July 19, 2011

    Stacey makes a very good point that the complexity to achieve scaling production systems is anything but trivial. Complexity driven by volume, concurrent users, and performance requirements of today’s new systems goes certainly way beyond the architectures of the OldSQL systems designed decades ago. However, few companies will be able to afford the amount of custom development a Facebook is doing on a daily basis to keep their huge,
    750+ million user base online. Clearly, the debate of what type of database system will serve next generation web and enterprise applications with “NoSQL” type requirements, such as linear scale on cost effective commodity hardware with no service interruption (high availability) is just getting started. Especially in enterprise applications, that typically deal with much more data complexity, data concurrency and data robustness, some of today’s biggest data demands are being met by object databases, a type of NoSQL database, think about it as a Key:Graph store. Here is an interesting case study from Sabre, serving large scale web sites such as
    Travelocity: http://www.versant.com/solution/Sabre.aspx

    Share

Comments have been disabled for this post