A Key To Cloud Standards: The Cloud Database

There has been a huge amount of talk about the need for standards in the cloud world. We recently joined the Open Cloud Manifesto in an effort to continue the dialog on the topic.

One of the key issues around establishing an Open Cloud is figuring out what should be standardized. To us, the key is to allow customers the ability to move applications from one cloud to another seamlessly. One problem inherent with cloud hosting is that it is, by definition, highly productized. So each provider will make choices that make moving somewhat complicated. But, these choices are also important for providing customers options in the cloud (for example persistent storage on our Cloud Servers offer vs. ephemeral storage on EC2). There needs to be a balance between ease of migration and product features.

One area we believe is headed in the wrong direction is the scalable database. The world already has two production-ready, next-generation databases: Bigtable and Dynamo. Amazon makes heavy internal use of Dynamo, but it is not available to the public (although SimpleDB may be based on Dynamo technology). Google offers access to Bigtable via App Engine, but doing so locks you in to their platform. Period.

Rackspace is strongly committed to helping establish a standards-based, next-generation database. And, we are putting our money where our mouth is.

One project that offers an open, scalable database is Cassandra. Originally developed by Facebook, the project is now part of the Apache foundation. We have a team devoted to the project and just last week, our team, lead by Jonathan Ellis, helped package up the first release candidate of the project. His discussion of the technology and this release is here. This is a first step for very bleeding-edge technology. We see great promise in the technology and would love to see broader engagement from other leaders in the cloud and developer community.

In the web era, scalable data stores are essential. But, the world needs standards in order to advance quickly. The relational database world standardized early, beginning with ANSI SQL-86 and -89. The results have been tremendous: customers can choose from many implementations, such as MySQL, SQL server and Oracle, without substantial lock-in. We hope to help a movement around standards of this type in the next-generation data store world. A set of technologies you can run with us, with one of our competitors, or on your own. Join us in support of these projects.

Lew Moorman is a senior consultant to the top executives of Rackspace, focusing on strategy and product issues. He also serves as a member of the Board of Directors. Lew joined Rackspace in April of 2000 and has served in a variety of roles, including as President and Chief Strategy Officer, while the company grew to $1.3 billion in annual sales. Before joining Rackspace, he worked for the consulting firm McKinsey & Company, advising technology clients on strategic issues. A native of San Antonio, Lew received a B.A. from Duke University and a J.D. from Stanford Law School.


  1. What about other scalable databases like:
    – CouchDB (which is also from Apache). this one is based on “documents”, and has ACID support.
    – Hypertable // this one resembles some “normal” relational features too, making it easier to use for a beginner. I think it has ACID support.
    – Voldemort: http://project-voldemort.com/ // only heard of it
    – Hbase // uses keys and values.

    From this list, I currently am experimenting with Hypertable, and CouchDB.

  2. Hey Lew,

    Cassandra makes a lot of sense as a unified database offering that can be provided across the many programming languages that Mosso supports (.NET, PHP, Python). However, are you looking into a solution to address the lion share of customers that are running on top of MySQL in the Mosso environment?

    Might I play matchmaker with Mosso and a small company called ScaleDB? They have shared disk solution similar to Oracle RAC at 5000 ft. As a system wide upgrade to Mosso’s current MySQL infrastructure it could be just what the doctor ordered. ScaleDB is a drop in storage engine that works with shared disk accessible via a SAN. ScaleDB would allow you to keep a similar model for pods that you have now, except you could scale successful clients to private pods like you do now except the node count could be 20 (master/master) instead of 2(master/slave).

    Looks like they had a pretty impressive showing at MySQL Con

    Check them out at http://scaledb.com


    Another option is to add SAN mounts (can grow into the terabyte range) as a feature for CloudServers so that individual customers can start leveraging this technology immediately. Ultimately, the best option is to adopt ScaleDB to implement the first CloudScale MySQL environment on the market. Then you could make optimal use of scale. In particular they have the concept of scaling in, which involves taking advantage of 64bit many core architectures running 8-32 instances of MySQL depending on the amount of dedicated cores.



  3. sorry for the delay in commenting.

    shahan312: we have looked at and used all of the mentioned tools. right now we see the most promise in cassandra. stay tuned for more details on why.

    travell: great point. 99% of the world’s applications will continue to be a great fit for relational databases and we continue to want to support broadly used standardized tools. as a provider that aims to make this all seamless and easy we have to help push the scaling of these tools. and we are. thanks for the tip on scaledb. we have engineers looking at many alternatives.


Please enter your comment!
Please enter your name here