Disks are currently 36GB 10kRPM SCSI ultra3 with hardware RAID (PERC 3/DI 
onboard with 128MB cache)

iostat for the DB disk partition is:

avg-cpu:  %user   %nice    %sys   %idle
           8.62    0.03    4.62   86.73

Device:    rrqm/s wrqm/s   r/s   w/s  rsec/s  wsec/s    rkB/s    wkB/s avgrq-sz 
avgqu-sz   await  svctm  %util
/dev/sdb1    2.74   4.63  8.58  7.10   90.54   94.58    45.27    47.29    11.81     
0.05    4.73   4.10   0.64

I'd say the 95-percentile for load is rather more like 1.4 than 2.

Thanks for the input,

A


On Mon, 16 Jun 2003, walt wrote:

> Andy Stubbs wrote:
> > 
> > Hi, long time listener, first time caller (I think),
> > 
> > I've got a database which I'm looking to increase performance, either by
> > buying bigger kit or by somehow optimising current configuration.
> > 
> > I'm running MySQL-Max-3.23.56-1 from the mysql.com RPMs on a Dell
> > Poweredge 2500 with dual 1400MHz PIII processors and 4GB RAM with RedHat
> > 7.3 kernel 2.4.18-10smp.
> > 
> > The database itself is actually quite small; about 3.3GB on disk. Disk
> > configuration is RAID-5, 3 disks, chunksize 8KB, default mounting options.
> > 
> > Currently, when moderately busy, it's not unusual to see 400 queries/sec,
> > so I imagine when running top whack at the moment we're probably hitting
> > up to 500 selects/second. Keeping our developers focussed on how their
> > design decisions affect performance is a continual process of course...
> > 
> > Anyway, the load average on this server is hitting 2 occasionally, and
> > it's time either to tune the configuration, move it onto some other kit,
> > or buy in some kit specifically. Apparently I might be looking at the
> > traffic on the database quadrupling in the next few months, so I'm keen on
> > getting this sorted ASAP.
> > 
> > Having Read The Fine Manual, and being more of a coder/sysadmin than a DBA
> > I have some dumb questions. Which are:
> > 
> > 1. Does anybody else have any experience with this situation? Good, Bad,
> >    Ugly?
> > 
> > 2. I can move this DB to a dual PE2600 with dual Xeon 1.8GHz processors.
> >    and I can get up to 6GB RAM in there. Is it worth doing this? i.e., how
> >    much extra capacity does this buy me? Does enabling HyperThreading on
> >    the Xeons help or hinder database servers?
> > 
> > 3. Would upgrading to 4.0.13 help at all? What kind of performance does it
> >    have compared with 3.23.56? This is a medium term goal anyway, and
> >    we'd like to take rather more time over it.
> > 
> > 4. I read that on a 32 bit architecture (like these Pentium class CPUs)
> >    the database tables are not memory-mapped (i.e., it's not possible to
> >    store the entire database in memory anyway). Is this the case with the
> >    64 bit Solaris too? Is there a planned implementation schedule for this
> >    functionality?
> > 
> > 5. What's the performance of MySQL like on Solaris 8/9 compared to Linux
> >    2.4.18? Is a big multiprocessor Sun box (like a Sun Fire 880 or 1280) a
> >    good choice for a database server running MySQL? Or are there more
> >    suitable platforms? Is, in fact, something like a SunFire 880 overkill?
> > 
> > 6. Would it be complete lunacy, in the absence of memory-mapped tables, to
> >    specify a RAM-disk on which to store the database? Should improve seek
> >    times, eh? And lots of redundant UPS stuff.
> > 
> > 7. Any other suggestions welcome. I'm a bit nervous about turning off
> >    atime on mounting the disks - is it really not used by the database
> >    anywhere? what kind of performance boost does it give?
> > 
> > Regards,
> > 
> > Andy
> > 
> > --
> > Andy Stubbs, B.A., Ph.D.
> > Network Manager, Active Hotels Ltd.
> > +44 1223 578106
> > 
> Andy,
> A load avg of 2 sounds like the machine is under a high I/O load.
> Have you considered using 15k rpm drives? Is your raid setup hardware or
> software?
> 
> walt
> 
> 

-- 
Andy Stubbs, B.A., Ph.D.
Network Manager, Active Hotels Ltd.
+44 1223 578106



-- 
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:    http://lists.mysql.com/[EMAIL PROTECTED]

Reply via email to