[ic] caching search results, academic discussion

[Jeff Dafoe]

Hello,

    I have noticed that, on my system, most of the load is generated by
product scans.  It seems to be the IO issue created when dozens of people
hit the first page of the search results at once and the system has to scan
the products and write the resultant match list out to the session
directory.  I suspect it is the writing of the results that creates the
highest load, as write operations are typically much slower than read
operations.  When pondering this, I realized that all of my users are doing
the same searches (via the category_vertical links) but each search is
resulting in redundant scans and writes.  I think there may be some
opportunity for caching here.  I am thinking it might be possible to create
one copy of the results that were returned for a given search instead of
each user having their own copy.  This global result list could be updated
at certain intervals, much like the timed-build tag.
    This is assuming I haven't missed some way to cache results page that
currently exists, which is not impossible.

Thanks,
Jeff