Re: [Sks-devel] web performance improvement

[Yaron M. Minsky]

The problem here is that the database is presently single-threaded, and
reinserting 200 keys requires a fair amount of work in terms of reading
the database and seeking all over the disk (the actual computational
component is pretty trivial).  It's not clear what the point would be of
making the web interface "live", since there's nothing it can actually
do.  Usually, people serve the query page off of a regular (and
separate) web server like Apache, and that's really the only thing that
can be broken off sensibly.

Now, you could try to allow for concurrent database access, which would
stop the database from ever appearing "dead", and would allow some
queries to proceed while the database was being updated.  That's
possible, but would require some non-trivial work.

Yaron

(btw, I do believe the SKS' performance is comparable to PKS'.  I'd be
interested in performance comparisons, though.)

On Fri, 2003-12-05 at 04:29, Fabio Massimo Di Nitto wrote:
> Hi Yaron,
>       as i wrote in one of my previous emails i am stressing sks quite a
> bit. This morning i did a farly simple load test.
> Grabbed 200 keys and reinjected them to the server (= no changes).
> while processing the keys at reinjection i did try to access the web
> interface but it was almost "dead" until sks completed the processing of
> the keys.
> 
> I think that perhaps having the web server on a separate process from sb
> could help in this case (if possible at all).
> 
> Any idea??
> 
> Thanks a lot
> Fabio
> 
> PS the box is quite powerfull but if you want tech specs please let me
> know.
-- 
|--------/            Yaron M. Minsky              \--------|
|--------\ http://www.cs.cornell.edu/home/yminsky/ /--------|

Open PGP --- KeyID B1FFD916
Fingerprint: 5BF6 83E1 0CE3 1043 95D8 F8D5 9F12 B3A9 B1FF D916