Jason,

Where did you get the Red Hat specifics from in your post?  Do you have
a reference document you could share?

Regards,

Paul Esson

-----Original Message-----
From: Justin Piszcz [mailto:jpiszcz AT lucidpixels DOT com] 
Sent: 01 May 2008 09:45
To: Esson, Paul
Cc: Jason Slagle; veritas-bu AT mailman.eng.auburn DOT edu
Subject: RE: [Veritas-bu] Anyone have any tweaks for LTO-4 and emulex
cardsunder RHEL5?

http://h71028.www7.hp.com/ERC/downloads/5982-9971EN.pdf

I think this is it-- it may have been LTO-2 and not LTO-3:

Results of HP testing with HP LTO2 tape drives
  Table 11.
  Number of tape transfer buffers    Number of waits Transfer rate to
tape 
(MB/sec)
  16                                 11,000          15
  24                                 3,000           16-17
  32                                 0               30-32
  48                                 0               30-32


On Thu, 1 May 2008, Justin Piszcz wrote:

> Unfortunately not it was from 2003 regrding the new LTO-3 (when it was
new) 
> on HP's site, it was a great review/document.
>
>> 16 for number_Data_buffers helped to 32, but after that it showed no 
> improvemnet and 256k was best size
>
>
> On Thu, 1 May 2008, Esson, Paul wrote:
>
>> Justin,
>> 
>> Do you have a link or other reference to the HP doc?
>> 
>> Regards,
>> 
>> Paul Esson
>> 
>> -----Original Message-----
>> From: Justin Piszcz [mailto:jpiszcz AT lucidpixels DOT com]
>> Sent: 30 April 2008 22:38
>> To: Jason Slagle
>> Cc: veritas-bu AT mailman.eng.auburn DOT edu
>> Subject: Re: [Veritas-bu] Anyone have any tweaks for LTO-4 and emulex
>> cardsunder RHEL5?
>> 
>> Btw there was a good HP doc, 32 for the NUMBER_DATA_BUFFERS and 256k
for
>> 
>> the SIZE_DATA_BUFFERS is the best.
>> 
>>> 32 (48) was not any better, but 32 > 16 was a definite gain in
>> performance.
>> 
>> 
>> Justin.
>> 
>> On Wed, 30 Apr 2008, Jason Slagle wrote:
>> 
>>> 
>>> Having just gone through that (Albiet with Qlogic) this is what I
did:
>>> 
>>> Make sure the st driver is set to a reasonable buffer size (I did
this
>> in
>>> /etc/modprobe.conf):
>>> options st buffer_kbs=1024
>>> 
>>> echo "1048576" >/usr/openv/netbackup/db/config/SIZE_DATA_BUFFERS
>>> echo "1048576"
>/usr/openv/netbackup/db/config/SIZE_DATA_BUFFERS_DISK
>>> echo "16" >/usr/openv/netbackup/db/config/NUMBER_DATA_BUFFERS
>>> echo "16" >/usr/openv/netbackup/db/config/NUMBER_DATA_BUFFERS_DISK
>>> 
>>> echo "kernel.shmmax = 268435456" >>/etc/sysctl.conf
>>> 
>>> echo "65536" >/usr/openv/netbackup/NET_BUFFER_SZ
>>> 
>>> Doing this, I am able to hit 70-80MB/s to the tape from my DSSU, and
I
>> think
>>> I COULD actually go faster, but my DSSU is limiting it.
>>> 
>>> Someone else amy have better values for these to improve it more - I
>> would
>>> love to hear them also.
>>> 
>>> Jason
>>> 
>>> --
>>> Jason Slagle
>>> /"\ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
>> . .
>>> \ /   ASCII Ribbon Campaign  .
>>> X  - NO HTML/RTF in e-mail  .
>>> / \ - NO Word docs in e-mail .
>>> 
>>> 
>>> On Wed, 30 Apr 2008, Justin Piszcz wrote:
>>> 
>>>> I am only seeing 8-10MiB/s directly from RAM (/dev/shm) where if I
>> write
>>>> the same data to an LTO-3, I see regular speeds, 60-90MiB/s no
>> problems?
>>>> _______________________________________________
>>>> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
>>>> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>>>> 
>>> 
>> 
>

______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email 
______________________________________________________________________

_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu