Start a Conversation

Unsolved

This post is more than 5 years old

W

465

February 21st, 2008 21:00

Clariion CX-300i slow reads - writes ok

We recently implemented a Clariion cx-300i (FLARE 02.24.301.5.016)and during some testing I've noticed that copies of a fairly large file (330MB) FROM the SAN to a user PC (through a server) is very slow, while writing the same file from the same client to the same LUN is reasonably speedy. No dirty pages shown, low SP util, read-cache is enabled. Does anyone have suggestions as to why this might be? Details are below:

Throughput copying from client to SAN is about 46Mbps
Throughput copying from SAN to client only 2.5 to 4.5 Mbps
The LUN in question is being served by a Win2K3 DellPE2950 with dual Gbps Netxtreme nics. Server is running Microsoft iSCSI init with latest version of PowerPath running in unlicensed mode.

San to Server connection occurs on an isolated, SAN-only LAN through a Cisco Gigabit 3560 switch - all san to switch and switch to server connections (1-only) are running at 1Gbps fdx and showing no errors. SPA-0 serves the LUN and on the server, NIC-1 connects to the SAN LAN.

On the Server, NIC-0 connects to the Customer LAN (CLAN). The LUN exists as a single, properly offset (64 not 128) partition with, currently, a single User share on the partition.

The Server, on the CLAN side, connects to a Cisco 3750 backbone switch @ 1Gbps fdx, which is trunked to a workgroup switch @ 1Gbps fdx via dot1q trunk. From there it's directly to the user PC at 100Mbps fdx.

The user PC is a Dell Precision dual Xeon 3.2G running XP Pro with a NetXtreme 57xx NIC.

When observing NIC utilization on the server, for the user to SAN copy you see a smooth pulse of traffic hitting about 9% on both the SAN NIC and CLAN NIC, just what you would expect.

For SAN to user copy, the CLAN NIC you see a long (6-12 min) rise in low-level (.3 to .5%) of traffic, and on the SAN NIC, nothing and next to nothing (less than .1%) through pretty much the whole thing.

Any ideas anyone?

Message was edited by:
wolf89

133 Posts

February 22nd, 2008 01:00

If you got access to Navi Analyzer can you confirm what size IO you are writing to and reading from the LUN? Also how is your RG setup, what are the prefetch settings for the LUN and cache settings for the array/LUN?

46 Posts

February 22nd, 2008 07:00

We don't have Navi-analyzer currently. We're in the process of purchasing it....

thx

46 Posts

February 22nd, 2008 09:00

All,

Thanks for your help. I have found the solution to this problem.

The issue turned out to be a Cisco IOS problem in the workgroup switch. Turning off QOS on the workgroup switch (no mls qos) immediately took read transfer rates from 2 to 4 Mbps to 40 to 47 Mbps.

thx
No Events found!

Top