Re: [storage-discuss] srp targe driver parameters

2010-11-21 Thread Susan Gleeson

No, there's no way to change those parameters at this time.

What problem are you trying to solve?

Giovanni wrote:

Hello,
via mdb is possible to verofy srp target parameters.

 target port: ff0945be9978
srq handle: ff0946a4da78
current srq size: 4095
max srq size: 4095
iu pool: ff09480c6000
profile send qdepth: 128
profile rmda read qdepth: 16
profile send msg size: 4160 
profile rmda xfer size: 65536

profile opcap mask: --x-x-xx

How is possbile to modify the send qdepth, read qdepth and msg size ?
Is there a config file or a command toold and which params need to be used ?

Thanks

___
storage-discuss mailing list
storage-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/storage-discuss


Re: [storage-discuss] srp targe driver parameters

2010-11-21 Thread Giovanni
Hi, I am trying to get max performance in sequential reads.

I did a lot of tests from a client (VSphere 4.1 + MLX OFED) tuning the 
parameters to get the max throughput from within a single VM.

I ended up getting IOmeter close to 2970 MB/s read (8MB requential block 100% 
read), but the standard throughput - as per people at MLX, should be 3200 MB/s 
(40 Gbit - 20% protocol overhead).

I am wondering where to further look.

On the client side the parameters I can work are:

ib_srp (srp client kernel module): srp_cmd_per_lun, srp_sg_tablesize, 
srp_can_queue

VMware Level: Disk.SchedQuantum, Disk.SchedQControlSeqReq, 
Disk.UserDeviceReset,Disk.SchedQControlVMSwitches, RequOutstanding
-- 
This message posted from opensolaris.org
___
storage-discuss mailing list
storage-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/storage-discuss


Re: [storage-discuss] srp targe driver parameters

2010-11-21 Thread Giovanni
As far as I can understand the current SRP target configuration seems to have a 
queue of 16 outstanding commands per target, is this correct ? After that a 
QFULL should be sent.

While at adapter port level the queue is 128, right ?

If am am not missing the point the target will be issuing a QFULL condition if 
overall queue commands 128 or single target LUN commands 16, right ?
-- 
This message posted from opensolaris.org
___
storage-discuss mailing list
storage-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/storage-discuss