I need to quickly replace a listings website which has the following characteristics:
- smallish database (10,000 items, < 1 GB)
- < 10% of the items updated/created/removed daily
- most common activity is searching the whole dataset, returning 1-1000 items
- traffic peaks at 1 million page impressions per day
The scaling strategy for the existing application has been to separate read-only and read/write activity. Multiple slave databases are used for searching and writes are done to a master, which update the slaves using MS SQL Server replication.
Since read activity is more common than write, this has proved to be a cheap way to do database load balancing, without true clustering.
I now need to replace the application – are there any C# open-source applications which scale as neatly as this?
As I see it searching for data and storing data are two different business operations. Personally I am a big fan of Lucene.NET for searching and would recommending investigating if it fits your needs. Lucene.NET has a number of pitfalls and writing to the index can be rather slow. In your situation I think you should stick with storing your data in a relational database (I.E., SQL Server) and look at modifying the searching component of your application.