Performance Problem between rp8400 and EMC Symmetrix 4.x

[Luigi Berengan]

I have a rp8400 hpux 11i connected via S.A.N (Brocade silkworm 2800) to a storage EMC Symmetrix 4.x with a lot of cache (I don't know if 10 or 12 Gb)with 7200 rpm disks.
When I copy a file (500Mb) from a logical volume to another on different LUN's, the time of cp commands is variable from 5 sec to 18 min for the same task.
Can someone tell me why and give me the problem solution ?

[Luigi Berengan]

I give another information : The HBA are two A6795A
Thank You

[Ashwani Kashyap]

Interesting problem .
The 6795A HBA's are 2Gb fibre cards and the silkworm 2800 is a 1Gb switch .

The HBA are set at auto neg by default . I think there might me a problem in the auto negotiating.

While the Auto neg on the HBA side work fine , I have noticed that the autoneg on the EMC FA's do not work fine all the time .

Ensure that the the FAs to which these HBA's are zoned is actually 1Gb (Hard code it ) and so is the switch .

[Luigi Berengan]

We verified. The FAs are set to 1Gbit.
Thenk you

[Michael Steele_2]

Start looking in the HBA's. Note total inputs and outputs in 'fsmsutil' for if your NOT load balancing then 95% of every transaction will be going through the primary pvlink.

A disk layout mapping through LUNS, via HBAS and into the EMC disk array is worth the week spent creating.

Do you have 'EMC ControlCenter' or 'SYMCLI'? These will help in the mapping and also provide reads and writes of each EMC device. (* Should have it.*)

Here's the server HBA side.

# ioscan -fnkC fc (* Note fc devices files, i.e., /dev/td1 *)

# fcmsutil /dev/td1

[Luigi Berengan]

The hosts are still in test.
There is no load on the hba. The only traffic are our tests of performance. With this results is impossible go in production.

[Stuart Abramson_2]

As stated above, map out the complete path that you take in copying:

Source:
HV (and Disk) -> DA -> FA -> FA Switch Port -> HBA switch Port -> HBA

Target:
Same as above.

Look for conflicts at each point. The best case is no duplication at all: i.e. - different HVs on different Disks (check all the disks, M1/M2, etc.) different DA paths, different FAs, differnet HBAs.

(The switch ports aren't so key here, unless you have a zoning problem of some kind.)

Do you have PowerPath installed? Are you sure that you are alternating PV paths on alternate links?

You check the HVs and Disks with:
symdev list
sympd list
symdisk list

I'm going to submit a disk copy ticket myself, watch for it...