[Cosmo-dev] Results of Hibernate Propfind initial testing

[Bobby Rullo]

After reading Randy's post with his results, I had a suspicion that  
my test was a little wack. His stuff took seconds to run, mine took  
many minutes.

I suspected that it might be the perl Http:DAV library that was slow  
- trying to parse the results of the propfind and whatnot. So I re- 
wrote the test using the underlying LWP useragent and the results  
were dramatically different...for the better!

Here's the new results:

Cosmo JCR, build: 2206
----------------------------------
Avg Time: 20 seconds

Cosmo Hibernate, build: 922
---------------------------------------
Avg Time: 2 seconds

A full order of magnitude better! THAT'S what we want to see!

Kudos and thanks to Randy for his great work on this and to bcm for  
driving this.

I'd like try this against events next....


Bobby


On Sep 2, 2006, at 12:17 PM, Brian Moseley wrote:

> On 9/2/06, rletness at simdeskcorp.com <rletness at simdeskcorp.com> wrote:
>
>> I haven't configured the l2 cache yet, so there is really no l2  
>> caching
>> now.  Thats a todo...and there are a lot of things to worry about  
>> when
>> determining what and what not to cache.
>
> cool. we'll definitely get some improvements there.
>
>> I happen to to have the dao layer timing for hibernate.  It takes  
>> around
>> the same time (7s) to find the collection by name, get all the  
>> children,
>> and retrieve the attributes.  This is MySQL again.  So the  
>> majority of time
>> is spent on the two queries that retrieve thousands of records.
>
> it would be interesting to record how often writes occur to the item
> and attribute tables vs reads. it's possible that mysql's query
> caching could help, if writes wind up being very infrequent.
> _______________________________________________
> cosmo-dev mailing list
> cosmo-dev at lists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-dev