Re: learning memcached

[Anonymous Monk]

in reply to learning memcached

What am I doing wrong, or are my expectations wrong?

I think you benchmark is flawed. You say you have 20_000 random entries, but you only make 2723 queries.

Instead of benchmarking for 3 wallclock seconds (the default, -3), try benchmarking multiples of 20_000 iterations.

After that, I would pre-load entries into the cache, and benchmark retrieval only.

[punkish]

> I think you benchmark is flawed. You say you have 20_000 random 
> entries, but you only make 2723 queries.

How did you determine that I made only 2723 queries? The benchmark shows that I made 40_000 queries for each code fragment at a rate of 2723 queries per second with the memcache process and 3846 queries per second with the straight-from-db process. Or, am I reading the results wrong?

Update: Here is another measurement, this time using timethese. The results are the same as before.

 query_dbh: 24 wallclock secs ( 5.28 usr +  5.29 sys = 10.57 CPU) @ 37
+84.30/s (n=40000)
 query_mem: 70 wallclock secs ( 7.80 usr +  6.08 sys = 13.88 CPU) @ 28
+81.84/s (n=40000)

Whatever I am doing, it is far faster to open up the SQLite database every time and query it than it is to query the memory cached value. (I am creating a $dbh and $sth every time, and also opening a file and writing to it every time to kinda emulate a CGI process that starts afresh on every hit of the browser.

[Anonymous Monk]

Or, am I reading the results wrong?

You're not reading the results wrong, I obviously missed the count :)

Another (insert better word here) "flaw" , is your sql query is too simple :) The way I understand memcached, you get benefit if you cache data which is expensive to calculate

[punkish]

> Another (insert better word here) "flaw" , is your sql query is too simple :) 
> The way I understand memcached, you get benefit if you cache data which is 
> expensive to calculate

I don't see why that should matter. Sure, on a more complicated and slow query, retrieving an already calculated result will be faster. However, retrieving even the simplest of results from memory should be faster than opening a file on disk, preparing a db handle, querying, and then returning the result. At worst, it should be about the same, not 30% worse!