Hello Simon,

Thanks for your reply,
The controllers are working in write-back mode and the battery voltage 
seems fine.
When manual writing a big file with dd and reading (copying) this file, 
the speed seems to be all right.
Here are some test results (DB1 = DELL and DB4 = HP).
Using bs=8k ore 16k ore something else doesn't make any difference.

Regards,
Bart Lubberdink

[EMAIL PROTECTED] backup]# time dd if=/dev/zero of=testfile.tmp bs=8k 
count=300000
300000+0 records in
300000+0 records out

real    0m28.807s
user    0m0.140s
sys     0m10.490s
[EMAIL PROTECTED] backup]#
[EMAIL PROTECTED] backup]# time cp testfile.tmp /database/

real    0m51.897s
user    0m0.190s
sys     0m15.170s



[EMAIL PROTECTED] backup]# time dd if=/dev/zero of=testfile.tmp bs=8k 
count=300000
300000+0 records in
300000+0 records out

real    0m40.721s
user    0m0.120s
sys     0m10.780s
[EMAIL PROTECTED] backup]# time cp testfile.tmp /database/

real    1m0.744s
user    0m0.170s
sys     0m14.490s





"Simon Matter" <[EMAIL PROTECTED]>
18-11-2005 13:00

                Aan:    [EMAIL PROTECTED]
                cc:     maxdb@lists.mysql.com
Onderwerp:      Re: Betr.: Re: Very poor performance of sapdb


> Hello Simon,
>
> Both raid controllers are brand new and working with cache.

Well, brand new doesn't mean the cache is working in write-back mode. If
the battery voltage is too low, the controller puts the cache in
write-trough mode until the capacity of the batteries is on the safe side.
So, really check this before going on, most controllers report this while
initializing.

>
> We use RedHat AS3 and when looking with the dset utility, the DELL 
server
> transferes aprox 35M/s read and write from/to the disk when restoring
> The HP does aprox 10k/s read/write!!
> When building the data volumes both servers worked fine.
> Both servers use 8G of memory.
> How can the CPU get involved in this? the CPU is completely bounced
> (95-100% I/O wait).

I have all kinds of DELL and HP servers here and they usually have similar
performance. Indeed, the HP raidcontroller is a bit better than the DELL
onboard ones.
Maybe you should first use something like bonnie to check the speed of
your filesystems.
And then when you check the db speed, are you sure both instances are
configured the same way? There are quite a number of parameters to check
here.

Simon

>
> Regards,
> Bart Lubberdink
>
>
>
>
> Hello,
>>
>> We use sapdb 7.4.3.31-1 on two different servers.
>> One server is a Dell PowerEdge 6850 and the other is a HP Proliant 
DL580
>> G3.
>> When checking both systems, the dell is a little bit faster when 
writing
>> to disk and reading from disk but this is normal because the Dell has
> 256M
>> cach on the raid controller and the HP 'only' 192M.
>> When i create a database from a backup, both machines create the log 
and
>> data volume almost in the same time (normal write speed)
>> There is one Log volume from 5G and 3 data volumes of 25G.
>> This takes about 40 minutes to build.
>> The problem begins when the actual data is imported form the backup to
> the
>> database.
>> The Dell server completes the restore (42G) in 40 minutes but the HP is
> so
>> very slow that after 3 hours, just 1G is imported!
>>
>> Is there a problem with the compatability between sapdb and the HP
>> controller "Smart Array 6i"?
>> Or does anyone have a clue what can be the problem here?
>
> There is much more involved than the RAID controller here. CPU, RAM, OS,
> almost everything. However, if you want to concentrate on the RAID
> controller, have a closer look at the cache. Do they both have battery
> backed cache, and are the batteries in good condition. Usually only then
> the cache is operated in write-back mode.
>
> Simon
>
>>
>> Regards,
>>
>> Bart Lubberdink
>> RAM Mobile Data
>>
>>
>
> --
> MaxDB Discussion Mailing List
> For list archives: http://lists.mysql.com/maxdb
> To unsubscribe:    http://lists.mysql.com/[EMAIL PROTECTED]
>
>
>
>


Reply via email to