No I don't know about an open benchmark, I have my own stuff for each project and I would recommend you to model your own. There are so many factors depending on what your application does that it wouldn't be meaningful to have a reference benchmark, it varies a lot depending on what you need and how you do it, how much you index, you store, the average size per document, the complexity of analyzers you choose.
I'll "shoot some numbers" to give you an idea: a simple fulltext query will usually return values in <10ms, linearly scalable across processors. So in worst case of all queries ==10ms - assuming no other processing is done by the same CPU - you can execute around 100 queries/sec/CPU On big indexes and when using complex queries (functions as "more like this") it might take longer, but it's unusual.
Adding stuff to the index is extremely unpredictable, I can't promise anything there, I can just tell you that my dual core laptop is able to add around 9000 documents/sec to the index, reading them from database on same machine (each "document" is a bunch of metadatas, around 30 fields not longer than a two-line description).
_________________ Sanne http://in.relation.to/
|