Sunday, 7 June 2009

New faster search for folia

Hey Everyone!

In this release we’ve change the search on the site around a bit. I’m sure you’ve all noticed how slow it was in returning a specific plant or journal post or even have slow it was in just saving posts and topics. Ultimately, the search that we had running wasn’t coping with the massive growth on the site. Our weekly usage stats keeps going up, everyone is journalling and gardening like mad and we have well over 100 000 different plants and varieties (most of which were added in the last few weeks).

Hopefully we’ve fixed all of this and sped things up with switching over to the new search. By our test runs its looking like a huge improvement. Indexing the site now take about 30 secs (down from hours and hours) and the results are retrieved in a faction of a second from the database and it shouldn’t be as slow when you save topics and posts.

However with all good things, there’s also some setbacks. Our old search was using a system called Ferret which was slow but allowed for updates in real time (so when you added a planting, it would instantly update the search) however we’ve switched to Sphinx which has quick indexing but no real time updates. To counter this – I’ve told Sphinx to reindex itself every 2 minutes or so. There’s a few other things I can do to make it feel more ‘instant’ but for now it’s a good starting point. Currently Sphinx is running on Plants, Varieties / Cultivars, Groups, Topics (yes! you can now search topics across groups! Posts are coming soon!) and Journals/Questions. So if something you added hasn’t popped up yet – give it some time and it will appear in a short while. Ferret is still used to index Gardens and Plantings as we need real time updates for the milestones and tasks – but we’re looking into switching over these too.

Also with the new search comes a different search algorithm – just like how Yahoo and Google never return the same results. So the results won’t appear exactly like they used to in the old search. These will be mainly for the better as we’ve got some good matching improvements, but there’s bound to be the odd plant that doesn’t appear in the list like it did before. It’s still very much a work in progress whilst we tweak it for the site.

So have a play around and tell us what you think.