MIT's Database System, Pyxis Makes Websites 3x Faster

MIT's engineers Alvin Cheung (first author), Owen Arden, Samuel Madden, Andrew C. Myers have proposed 'Pyxis' - a new system that streamlines the database access patterns to make the database driven websites and applications serve up to 3x faster. In their <a href="https://people.csail.mit.edu/akcheung/papers/vldb12.html" target="_blank" rel="noopener noreferrer">Automatic Partitioning of Database Applications</a> titled 'Automatic Partitioning Of Database Applications', the authors mention that their experiments with TPC-C and TPC-W revealed that the Pyxis system could result into 3x reduction in latency and 1.7x improvement in throughput compared to the traditional, unpartitioned systems. What Pyxis does is automatically partition the application code into two pieces; one executed on the application server and other on the database server. The system analyses and profiles the server loads and code dependencies and then uses the statistical data to create partitions that reduce the number of control transfers and the overall amount of data exchanged.



The team promises that the Pyxis system works well with the existing programming languages popular among the web developers and do not require learning of an entirely new language. Those who're interested in getting into the details of Pyxis, may refer to the <a href="https://people.csail.mit.edu/akcheung/papers/vldb12.pdf" target="_blank" rel="noopener noreferrer">PDF</a> (PDF).