From: Mounir STAMBOULI (mounir.stambouli@LAPOSTE.NET)
Date: Wed Feb 19 2003 - 15:43:59 EST
We have 4 FC with 2 pathes for each vpath. When running night jobs, wait-io
is low. Should it make a difference to distribute pathes over the 4 adapters
?
Mounir
----- Original Message -----
From: "Adams Kevin J" <kevin.adams@PHS.COM>
Newsgroups: bit.listserv.aix-l
To: <aix-l@Princeton.EDU>
Sent: Thursday, February 13, 2003 8:32 PM
Subject: Re: disk I/O throughput
> Mounir,
>
> Your fcs0 adapter settings for num_cmd_elems and max_xfer_size are at max,
> so not much you can do there.
>
> However, your hdisk queue_depth might be a little low, with the range
being
> 1-256. A redbook I once read suggessted 75, but I'm unable to find it to
> verify.
>
> We have a 2109 switched SAN, with 4 path vpaths and 2 6227 FC adapters per
> host, and according to NMON, we max out at 18 MBps per vpath. We little or
> no i/o wait though.
>
> We have your exact settings for the adapters and hdisk. We had to lower
the
> hdisk queue_depth back down from 75 to 20 because we were overrunning the
> host adapter num_cmd_elems, and this was done at ML08 without a PTF to
apply
> in emergency mode when the adapters froze. The temp fix was to raise the
> adapters num_cmd_elems to max, and lower hdisk queue_depth back to default
> of 20. It's been fixed via PTF, just can't make any more changes right
now.
>
> At a TSM class at Userblue, the speaker said they have found the host FC
> adapters to be the limiting bottleneck. You have only 2?
>
> Kevin Adams
>
>
> -----Original Message-----
> From: mounir.stambouli [mailto:mounir.stambouli@LAPOSTE.NET]
> Sent: Thursday, February 13, 2003 7:32 AM
> To: aix-l@Princeton.EDU
> Subject: [aix-l] disk I/O throughput
>
>
> Hello,
>
> We've got an AIX 5.1 box with FC adapters directly connected
> to ESS FC ports (no SAN). The theoretical throughput is 100+
> MBPS but nmon reports that we're getting no more than 8
> MBPS !!! I have paste the output of lsattr of various devices.
> ESS volumes are configured as RAID5 seen through 2 paths for
> each vpath.
>
> Is there any additional config to be done ?
> Where should I look on the ESS ?
>
> Thanks in advance,
> Mounir
>
>
>
> U62CTSM001:/ lsattr -El fcs0
> bus_intr_lvl 51 Bus interrupt
> level False
> intr_priority 3 Interrupt
> priority False
> bus_io_addr 0xec00 Bus I/O
> address False
> bus_mem_addr 0xe0020000 Bus memory
> address False
> lg_term_dma 0x200000
> N/A True
> max_xfer_size 0x1000000 Maximum Transfer
> Size True
> num_cmd_elems 1024 Maximum number of COMMANDS to queue
> to the adapter True
> pref_alpa 0x1 Preferred
> AL_PA True
> sw_fc_class 2 FC Class for
> Fabric True
> init_link al INIT Link
> flags True
> U62CTSM001:/ lsattr -El fscsi0
> scsi_id 0x1 Adapter SCSI ID False
> attach al How this adapter is CONNECTED False
> sw_fc_class 3 FC Class for Fabric True
> U62CTSM001:/ lsattr -El hdisk13
> scsi_id 0xef N/A
> True
> lun_id 0x5029000000000000 N/A
> True
> location N/A
> True
> ww_name 0x5005076300c68efe N/A
> False
> pvid none Physical volume identifier
> False
> q_type simple Queuing TYPE
> True
> queue_depth 20 Queue DEPTH
> True
> start_timeout 180
> eout 60 READ/WRITE time out value
> True
>
>
> Accédez au courrier électronique de La Poste : www.laposte.net ;
> 3615 LAPOSTENET (0,13 EUR/mn) ; tél : 08 92 68 13 50 (0,34EUR/mn)"
>
>
> This electronic message transmission, including any attachments, contains
information from PacifiCare Health Systems Inc. which may be confidential or
privileged. The information is intended to be for the use of the individual
or entity named above. If you are not the intended recipient, be aware that
any disclosure, copying, distribution or use of the contents of this
information is prohibited.
>
> If you have received this electronic transmission in error, please notify
the sender immediately by a "reply to sender only" message and destroy all
electronic and hard copies of the communication, including attachments.
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:16:36 EDT