List:Database Benchmarks« Previous MessageNext Message »
From:Peter Zaitsev Date:June 15 2004 6:07pm
Subject:Re: benchmarks-plain Digest 14 Jun 2004 22:30:01 -0000 Issue 48
View as plain text  
On Mon, 2004-06-14 at 23:36, datasmid@stripped wrote:
> > Two questions. Which Kernel did Fedora Core 2 run ?  Also did you use
> > hardware or software mirror ?
> >
> Linux fedora 2.6.5-1.358smp with hardware mirror

So you're mostly testing 2.6 kernel vs 2.4 not a big surprise.

> >>
> >> Conclusion: Fedora Core 2 shines brightly!!!
> >
> > Well, It was twice as slow to run ATIS, big-tables, connect, wisconsin.
> > It was also slower at "select"
> >
> > So results look more like "depends on the load".
> >
> 
> Well, we need the fastest inserts to record the financial markets, so if
> we improve total benchmark time with faster inserts we are happy.

Heh,

Do not be mistaken by the name. "Insert" is not very well defining name.
There are a lot of things in this benchmarks other than just inserts. 
If you take a look at insert operation itself:

insert (350768)                     |  36.00|  38.00|  66.00|
insert_duplicates (100000)          |   8.00|   8.00|  15.00|

As you can see it can be significantly worse.  It however also still
depends on number of keys you have and table size. Disk bound load is
whole different world. 

-- 
Peter Zaitsev, Senior Support Engineer
MySQL AB, www.mysql.com



Thread
Re: benchmarks-plain Digest 14 Jun 2004 22:30:01 -0000 Issue 48datasmid15 Jun
  • Re: benchmarks-plain Digest 14 Jun 2004 22:30:01 -0000 Issue 48Peter Zaitsev15 Jun