Enhancement of Middleware for Stanford WebFEATURE Project
Sonal P. Mahajan
Oral Defence Date:
Friday, May 11, 2012 - 16:30
Professors Dragutin Petkovic, James Wong, and CCLS Research Assoc. Mike Wong
In order to allow broader access to Stanford FEATURE application, a bioinformatics system for modeling functional sites in protein structures, a WWW site was created by Stanford Helix group and SFSU team, called WebFEATURE (http://feature.stanford.edu/webfeature/) which offers the FEATURE functionality as a" service". This site allows users to upload the PDB sequence file and get the result in the form of display of detected functional sites superimposed on molecular structure. Efficiency with regards to performance and security are important features of bioinformatics websites. The efficiency in terms of response time and reliability of existing, original WebFEATURE with regards to performance primarily relied on the job scheduler used in the middleware. Tests revealed that original WebFEATURE system reached a job success rate of 66.2% which was unsatisfactory. In addition, the original implementation of the WebFEATURE middleware was vulnerable in terms of security. It has been observed that these factors are having negative implications on the users, driving them away from using WebFEATURE. This motivated our work, which focused on the two main goals: a) Improving the efficiency of WebFEATURE, specifically the scheduling system to reduce response time by studying the existing middleware in WebFEATURE, analyze bottleneck and problem areas and design, implement and test necessary changes in consultation with Stanford Helix group; and b) Improving the security of WebFEATURE. The existing middleware concentrated around the scheduling system was studied first. Different available open source schedulers were researched according to the desired objectives to find the probable best fit into the WebFEATURE system. Out of the many, two candidate schedulers were identified and a prototype was implemented, out of which one, "APScheduler" was finalized, since it satisfied the requirements most effectively. The scheduling system of WebFEATURE was replaced with the help of this new scheduler, and various functional, load and performance tests were conducted to validate its working. The new scheduler was observed to be approximately 10 times faster than the one present in the existing version of WebFEATURE. It also increased the job success rate to about 99%. The new scheduling system was incorporated in the WebFEATURE 3.0 rollout and deployed on the Stanford server in February 2012. Security audit was conducted to find vulnerabilities of the WebFEATURE website. Identified high priority security issues were fixed to reduce the security risks and integrated into live version of WebFEATURE.
bioinformatics,job scheduler, security audit, APScheduler, WebFEATURE
Sonal P. Mahajan