[pmwiki-users] Cacheing pagelist results within the same page for duplicate searches?

[Peter Bowers]

On Fri, Aug 22, 2008 at 3:15 PM, Peter Bowers <pbowers at pobox.com> wrote:
> I'd be curious if anybody is actually doing benchmarks
> what the actual comparison is:
>   1. doing it 3 times without caching
>   2. doing it 3 times with caching
>   3. doing it using the name method as I'm doing

Now that Christian pointed out PageGenerationTime I went ahead and ran
some benchmarks.  I used test data of about 380 pages using a PTV
condition which matched about 180 of those pages.  My host is somewhat
slow which I would guess helps with benchmarking?  I ran each test 3
times for averaging.

NO SORT ON ANY PAGELIST (the 3 pagelists were identical in every way):
 using caching: 14.4 average
 not using caching: 20.7 average
 using the "name" method in combination with caching: 14.1 average
 using the "name" method without caching: 21.3 average

USING THE 3 DIFFERENT SORTS AS DEFINED (-ctime, -time, alpha):
 using caching: 17.0 average (5 second difference between the first
time and subsequent)
 not using caching: 20.8 average
 using the "name" method in combination with caching: 17.1 average
 using the "name" method without caching: 21.3 average

So caching makes a big difference (as much as 25% in some cases).
Using the "name method" is pretty pointless - maybe a 2-3% improvement
but sometimes actually worse.

FYI.

-Peter