Original URL: http://www.theregister.co.uk/2008/07/18/facebook_cassandra_microsoft/
Microsoft, Facebook, Google box clever on really big systems
The enemy of my enemy...
Posted in Applications, 18th July 2008 17:02 GMT
Watch Now : Virtual Machine Movement with Hyper-V
Facebook's decision to release under open source [1] a large-scale data management [2] project similar to - and inspired by - Google's BigTable [3] has received backing from an unusual quarter: Microsoft.
Data center futures architect and distinguished database developer James Hamilton [4], has complemented the pimply faced social network for releasing what he said [5] "looks like a well-engineered system."
Hamilton spent ten years at IBM working on DB2 before joining Microsoft in 1997 to work on SQL Server, and recently collaborated [6] with fellow database guru Michael Stonebraker on an examination of future database architectures.
Stonebraker put the cat among the pigeons by slagging off [7] Google's MapReduce [8] database tool earlier this year - he called it a step backwards. Stonebraker also noted there are limitations in BigTable and its open-source equivalent Hbase [9].
Conspiracy theorists might find it interesting that Stonebraker's co-author David de Witt joined [10] Microsoft in April to head up a new research effort into large databases. They should also remember Microsoft is an investor in Facebook.
Adding insult to injury, Facebook has put its project - called Cassandra - up on Google Code. Cassandra is not alone on Google code. Another BigTable clone called Hypertable [11] was set up on Google Code [12] earlier this year.
Like BigTable, Cassandra is designed to get round the limitations of traditional relational databases in large-scale, online applications.
Cassandra is the work of a Facebook team led by Jeff Hammerbacher [13], an ex-Harvard student who was recruited by Facebook chief Mark Zuckerberg from troubled Wall Street bank Bear Stearns.
Hammerbacher is now reported [14] to have become the latest Faceboook employee to have served his notice. Can we expect this Facebook youf to show up in well-remunerated style at Redmond this Fall?
And, will all this prompt Google to put BigTable - or even MapReduce - into open source? Given it decided to release its Protocol Buffers [15] technology to open source this week - its not beyond the bounds of possibility.
There is a precedent: Facebook released Thrift, its clone of Protocol Buffers, to open source last year.
The games continue.®
Links
- http://code.google.com/p/the-cassandra-project/
- http://www.slideshare.net/jhammerb/data-presentations-cassandra-sigmod/
- http://labs.google.com/papers/bigtable.html
- http://www.mvdirona.com/jrh/Work/
- http://perspectives.mvdirona.com/2008/07/12/FacebookReleasesCassandraAsOpenSource.aspx
- http://www.nowpublishers.com/product.aspx?product=DBS&doi=1900000002
- http://www.theregister.co.uk/2008/01/24/stonebraker_dewitt_mapreduce/
- http://labs.google.com/papers/mapreduce.html
- http://hadoop.apache.org/
- http://www.theregister.co.uk/2008/04/28/microsoft_hires_dewitt/
- http://www.hypertable.org/about.html
- http://code.google.com/p/hypertable/wiki/SourceCode?tm=4
- http://jeffhammerbacher.com/
- http://venturebeat.com/2008/07/10/rumor-jeff-hammerbacher-a-key-early-facebook-employee-is-leaving/
- http://www.theregister.co.uk/2008/07/14/dziuba_google_protocol_buffer/page2.html
