Re: [greenstone-users] Time taken

From George Buchanan
DateTue, 7 Oct 2003 09:41:51 +1300
Subject Re: [greenstone-users] Time taken
In-Reply-To (Pine-LNX-4-33-0310061705060-24024-100000-mmsl-serc-iisc-ernet-in)
> I need to check, How much "time" greenstone takes to fetch the
> result for the given query?
...this varies, depending on the individual query that is being

> I mean, i have to find the "Time Taken" between the 'query' given
> and the 'results' returned from the DataBase by the search engine.
...time taken at which point? If you mean from a client on a remote
machine, then clearly there are issues of network latency, etc. If
not, and you're only interested in the time taken on the server, then
clearly the hardware of the server would be a factor. Should it be
the case that you are interested in your own particular server's
performance, it may be worthwhile undertaking a number of tests.

Generally, Greenstone performs searches very rapidly - if you're
interested in the details of its search engine and its performance,
I'd recommend reading "Managing Gigabytes" by Ian H. Witten,
Alistair Moffat and Timothy C. Bell, published by Morgan
Kaufmann, ISBN 1558605703. The book describes the search
engine used by Greenstone, MG, and its technical construction in
some detail.

If you are wishing to compare the search speed with other
search engines, then I'd advise caution in your methodology -
different sized corpuses will perform differently on the same
search engine, and the decay of speed with corpus size requires
detailed consideration - furthermore there is a high element of
variability of the same search on the same computer under
most contemporary operating systems.

The one exception to search results being returned rapidly is when
one performs phrase searches using the original MG search engine.
If you require that facility, you would be better advised to reconf-
igure your collection to use MGPP, which natively supports phrase
searching and also supports fielded searching as well. If speed of
searching appears to be inadequate, then it probably is down to
using MG with phrase searching.

If, on the other hand, you want merely to present the time taken
to perform your query on the search results page, then that would
be best achieved by adding a new macro and setting the definition
of that macro in queryaction.cpp.

I'm afraid that without a bit more insight into why you feel it is
important to time the search, I can't be of more help.