FreBSD 9.1 and ZFS v28 performances

Steven Hartland killing at multiplay.co.uk
Tue Mar 19 14:05:23 UTC 2013


----- Original Message ----- 
From: "Davide D'Amico" 
> i don't know if you have installed sysbench 0.5.0 somewhere in you 
> server, but we could test using *standard* oltp tests (the dataset I'm 
> using is strictly private) included in sysbench package.
> 
> Is it possible, for you?
> 
> I'm using mysql-5.6.10 enterprise (but I think the one you find in the 
> ports tree it's a good choice, too).

I've installed sysbench latest version from ports which is 0.4.12_1.

Having run the standard oltp both read only and read + write it doesn't
seem like a very good benchmark for disk as it seems like it does
very little disk access.

== ZFS ==
Test execution summary:
    total time:                          248.5637s
    total number of events:              1000468
    total time taken by event execution: 3968.4764

== UFS ==
Test execution summary:
    total time:                          220.7406s
    total number of events:              1000588
    total time taken by event execution: 3523.5829

I noticed that UFS's support for TRIM is very poor compared to
ZFS's with it queuing all TRIMs till after the test completed as well 
as performing many more BIO_DELETE requests causing
significant system overhead and disk saturation after completion.

    Regards
    Steve

================================================
This e.mail is private and confidential between Multiplay (UK) Ltd. and the person or entity to whom it is addressed. In the event of misdirection, the recipient is prohibited from using, copying, printing or otherwise disseminating it or any information contained in it. 

In the event of misdirection, illegible or incomplete transmission please telephone +44 845 868 1337
or return the E.mail to postmaster at multiplay.co.uk.



More information about the freebsd-fs mailing list