DSS Performance

[brent192]

Im seeing poor performance on using DSS 1.32.DB00000000.2854.
Hardware is:

Tyan S3970G2NR -H1000E
IntelPro Gig Ethernet
3Ware 95550 SX Raid 50 - with write cache and storsav set to performance
Dual rev.f Opteron
4GB RAM

Im seeing a max of about 30Mb transfers using iscsi with Microsoft or Linux-iscsi. Any ideas for tweaking?

Thanks,

Brent

[To-M]

Was the 1.32 an upgrade from 1.23 or is this a new install with new Volumes from the 1.32? Is the 3Ware RAID firmware up to date. Are you bonding (which mode) if so can you test without the bond. Is the Target with WB enabled? Try testing with forcing the nic speed - in console screen enter ALT+CTRL+T->Modify driver options select Intel option and choose AutoNeg. Set this value to 32.

[brent192]

It is an upgrade from 1.23. I tested with the CD before we purchased your product, not sure the version, and was seeing much higher performance - 80+Mb. I am not bonding. The volume I am testing was created after the upgrade. How do I set WB on the target? I will update the software on the raid card and set the autoneg in the console. I will have to do this during a maintainence window. This is a production system. Thanks!

[To-M]

In release notes it states that when upgrading to this version it is recommended to remove LV's and then recreate them due to the Block IO - 1.23 is File IO. You will have to save your data. Although you don't have to.

* IMPORTANT : Since ver 1.30 New iSCSI default volume creation is done in block-IO in contrast to the older version that was file-IO.
In order to create new iSCSI volume in "old" file-IO mode, please switch the default in console tools menu: ctrl-alt-w --> Tuning options-->iSCSI daemon options-->iSCSI volume type.
Block-IO mode is about 30% faster then File-IO and the target volume size is exactly equal to defined iSCSI volume size (in file-IO the target size is a bit smaller then defined iSCSI volume size).
Additionally, initialization is no-longer required as with the old file-IO volumes.
In order to migrate your data from the file-IO to block-IO volume, the data must be backed up from the existing iSCSI volumes and then restored into newly created iSCSI volumes.
NOTE: please verify data integrity prior to deleting old iSCSI volumes.
Volume replication is only possible between to similar volumes (e.g. block-IO <-> block-IO or file-IO <-> file-IO). Volume replication between old file-IO and block-IO volumes are NOT possible.

[Raudi]

Bad performance with BlockIO (http://forum.open-e.com/showthread.php?t=562)

Perhaps, the same Problem? What happens when you access the storage with two clients at the same time?

I have a 16 channel 3ware 9550SXU with BBU in Performance mode and one Raid5 with 6x Seagate 750 ES.

Greetings
Stefan

[To-M]

In order to have more connections at the same time to the same Target you would require a special SAN file system. iSCSI Target works block based - not file based. Meaning that a changes made by one person is not seen by another person working on the same target/volume. Also this can cause issues with the data, below is a list of solutions to do this. Did you send in the logs and your tests from OpenFiler to support concerning your speed issues?

SAN file system solutions are very expensive. The easiest way to share the data on iSCSI Target is to create network share and have the clients connect to the network share and not directly to the target. In case clients read only from the target and need to access recent data (written from single host only) they can logout and login new session in the initiator. This will allow them to see the recent data.

To control shared iSCSI target you need so called SAN File System programs.

MelioFS:
http://www.sanbolic.com/melio.htm
IBM:
http://www-306.ibm.com/software/tivoli/products/sanergy/
http://www.ibm.qassociates.co.uk/storage-software-totalstorage-san-files
yste
m.htm
StorNext:
http://www.adic.com/ibeCCtpItmDspRte.jsp?section=10024&item=121889
SGI:
http://www.sgi.com/products/storage/tech/file_systems.html
PolyServe
http://www.polyserve.com/
DataPlow SFS
http://www.dataplow.com

[Raudi]

Not so complex! :rolleyes:

One RAID and one Volume Group. In this Volume Group you create 2 Logical Volumes so you have 2 targets. Now you can connect each client to a target and can access/test the RAID from two clients at the same time...

Greetings
Stefan

[brent192]

I already have multiple targets on one volume group. Perhaps this is when the performance issue began? To be exact - I have 2 targets - both have 3 lv's mounted.

Brent

[netsyphon]

I already have multiple targets on one volume group. Perhaps this is when the performance issue began? To be exact - I have 2 targets - both have 3 lv's mounted.

Brent


should be ok given you can accept that they may end up fighting for "king of the spindle".

I have a 4 disk raid 10 (SATA) with a 9650sx and I peak out at 112mb/s between the 2 virtual machines that are housed there.

I think tom is right, you didn't update properly to block i/o. thats going to change your performance by ~30%. if that still doesnt help then can you tell us what kind of switch you are connected to?

About hardcoding interfaces: Is that you never want to disable auto-neg on only 1 side - if you hardcode one side you need to go ahead and hardcode your switch as well.

you definitely want to update all your firmware on the mainboard, drives and controller card!

[brent192]

I have created a new target after the update with the same performance. If have tried two switches - Netgear GS724T and Allied Telesis 8724XL.