News

Database doctorate provides breakthrough for multiple users


4 February 2010

University of Sydney alumnus Dr Michael Cahill has received an Australasian award for outstanding doctoral work in the field of IT.

Dr Cahill won the 2010 Computing Research and Education Association of Australasia (CORE) Distinguished Doctoral Dissertation prize for his research into database management.

His PhD thesis entitled Serializable Isolation for Snapshot Databases, saw him develop an algorithm to avoid data inconsistency in databases accessed by multiple users.

The algorithm tackles problems occurring when many clients access a database simultaneously. While organisations like banks lock their databases so only one person can access a single piece of information at a time, this isn't tenable for retail websites like Amazon where thousands of people often want to the same book title at the same time.

Most large databases therefore opt for what's called a multiversion system. The existing multiversion algorithms work well most of the time but their potential to produce inconsistent data can have serious consequences, says Dr Cahill. He cites the example of doctors scheduled to work at the same time changing their rosters simultaneously so that no one is rostered at all.

"I looked at algorithms used in real systems and found a practical way to maintain consistency," he says. "I've changed the algorithm in multiversion systems so they keep track of more information. By tracking some additional information about concurrent operations, the algorithm guarantees the traditional kind of correctness expected of a database, or, to use the correct terminology, prevents anomalies."

The Distinguished Doctoral Dissertation prize was awarded to last month at Australasian Computer Science Week. Dr Cahill's work was previously been recognised at the 2008 ACM SIGMOD International Conference on Management of Data, winning the Best Paper Award. The paper was chosen as one of the core readings in the graduate Database course at University of California at Berkeley.

Dr Cahill's algorithm has been implemented in some open source systems and he is discussing its potential use with commercial vendors. He currently works for Oracle, a multinational computer technology corporation.
Media contacts: Jocelyn Prasad 02 9114 1382 / Sarah Stock 02 9114 0748 sarah.stock@sydney.edu.au