Big improvements to the BII web application (version 1.4)
A few weeks ago, we started loading the InnoMed datasets in the EBI instance of the BioInvestigation Index and we started to notice performance problems, especially when accessing the detailed study pages (it was taking almost 70 seconds it to render the pages). We found two problems: the growing number of annotations in the database (1.6 million) were increasing the query space considerably; and 2) the inefficiencies present in Hibernate queries resulted in larger scope queries than we really wanted .
So, we improved the Lucene index which was generated from the loader and already consumed by the web application to generate the browse view and made it capable of holding all of the assays, factors, characteristics, publications, contacts & their respective content so that we would remove the inefficient queries necessary to generate the study page and therefore make things quicker. Now, everything is quicker and there is no real latency in the page load process. Plus, we’ve also adding server-side caching to further reduce lag. So now, pages such as this one http://www.ebi.ac.uk/bioinvindex/study.seam?studyId=BII-S-8 load almost instantly, compared to the 90 seconds it took beforehand (obviously completely unacceptable).
If you visit the EBI or Harvard Stem Cell Discovery Engine instances of the BII, you’ll immediately see the improvements to the web application look and feel. We’ve made considerable improvements to clean up the interface and make browsing the content much easier.
We still have some plans to even further enhance the web application, so watch this space for version 1.5! You can download version 1.4 from here…you’ll need maven 2.2.1 and JBoss 5 installed first though! There are instructions on our GitHub pages about how to build!