How do I do a free text search in Azure Table Storage?

Question

I have a solution with a Azure table storage with up to a few tusands "rows" per customer (partition key).

How do I best do a lightning fast free text search?

Because of the nature of the data I'm not able to do a hole word search (eg. a search for "zur" should match "Azure").

Answer 1

Just spotted this which may help you: Azure Library for Lucene

Answer 2

We are using the following in production for our sites: We run hosted solr (based on lucene) instances on http://websolr.com and cache the results using the new azure distributed cache feature that is currently in beta. That gives us a worst case 200 ms latency for an initial search request between the Amazon datacenter where websolr.com runs and the Azure Datacenter and an average 6 - 10 ms for all cached searches. We also record common search text fragments and try to keep them fresh in the cache.

Answer 3

At the moment there is no out of the box solution for this. Perhaps a full text search feature will be announced at PDC10.

So at the moment you will need to roll your own text indexing solution. The way I have done this is by building a Lucene.net index on a worker role. I then open a tcp port on that worker role that provides a search service using WCF. Any web role can then consume that service. This works really well and provides a very fast search service.

There is a PDC09 video by Steve Marx that gives more information: http://www.microsoftpdc.com/2009/SVC16