I could spend a week or two tweaking the performance of my database servers
and probably make some sizeable improvements, but I'm not going to.

Why? Because PostgreSQL screams as it is.

I would make sure that if the consensus is to add some sort of caching that
it be done only if there is no hit to current performance and stability.
That being said, I think that server side caching has major buzz and there's
nothing wrong with adding features that sell.

I will disagree with 3 points made on the argument against caching.
Specifically, the benefit of doing caching on the db server is that the
benefits may be reaped by multiple clients where as caching on the client
side must be done by each client and may not be as effective.

So what if the caching has a slight chance of returning stale results?  Just
make sure people know about it in advance.  There are some things where
stale results are no big deal and if I can easily benefit from an aggressive
caching system, I will (and I do now with the adodb caching library, but
like I said, caching has to be done for each client).  In fact, I'm all for
using a low-tech cache expiration algorithm to keep complexity down.

Finally, if the caching is not likely to help (or may even hurt) simple
queries but is likely to help complex queries then fine, make sure people
know about it and let them decide if they can benefit. 

Sorry if I'm beating a dead horse or playing the devil's advocate.  Just
felt compelled to chime in.

Matthew Nuzum      + "Man was born free, and everywhere
www.bearfruit.org  :  he is in chains," Rousseau
+~~~~~~~~~~~~~~~~~~+ "Then you will know the truth, and 
the TRUTH will set you free," Jesus Christ (John 8:32 NIV)

-----Original Message-----
[mailto:[EMAIL PROTECTED] On Behalf Of Tom Lane
Sent: Monday, September 27, 2004 1:19 AM
To: Neil Conway
Cc: Aaron Werman; Scott Kirkwood; [EMAIL PROTECTED]
Subject: Re: [PERFORM] Caching of Queries

Neil Conway <[EMAIL PROTECTED]> writes:
> I think the conclusion of past discussions about this feature is that
> it's a bad idea. Last I checked, MySQL has to clear the *entire* query
> cache when a single DML statement modifying the table in question is
> issued.

Do they actually make a rigorous guarantee that the cached result is
still accurate when/if it is returned to the client?  (That's an honest
question --- I don't know how MySQL implements this.)

IIRC, in our past threads on this topic, it was suggested that if you
can tolerate not-necessarily-up-to-date results, you should be doing
this sort of caching on the client side and not in the DB server at all.
I wouldn't try that in a true "client" scenario, but when the DB client
is application-server middleware, it would make some sense to cache in
the application server.

                        regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
    (send "unregister YourEmailAddressHere" to [EMAIL PROTECTED])

---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
      joining column's datatypes do not match

Reply via email to