I have three pieces of 10GBE related feedback, which may be a bit tangential or redundant.
1) Don't assume that the average OS network stacks are going to make efficient use of or good throughput on 10GBE. Insofar as filers also run an OS, its not a given that they will perform at peak form either. (This may have changed since I worked with 10GBE three years ago, but I would document your assumptions, and document nominal throughput wherever you can just for the insight you'll gain.) 2) Don't assume that any two different models/brands of 10GBE NICs supplied by the same platform vendor for use with the same OS are going to perform alike. Also, don't assume that GA firmware isn't crap because its the only version available. 3) Look for an ugly knee in the throughput curve as you approach full scaleout of the 10GBE switches. I've seen at least one workgroup network switch that ran short on an internal resource as more than three quarters of the network ports were in use. Likely a thing of the past as well. [RC] From: James R Owen <[email protected]> To: [email protected] Date: 05/05/2010 04:37 AM Subject: [ADSM-L] ? linux TSM 6.2 using 10GE iSCSI ... to backup HPC ? Sent by: "ADSM: Dist Stor Manager" <[email protected]> ?Are you: doing TSM backups to iSCSI disk/tape? w/10GE? w/TSM6.? Scanning ADSM-L archives, actual TSM experience w/iSCSI seems scarce! Recently, Gary Bowers recommends "DirectIO No" in dsmserv.opt w/ iSCSI, and in 2007 Ben Bullock had iSCSI poor performance (w/o that setting?) Most other iSCSI threads are questions answered by theory/opinion. We hope to backup HPC (primarily Gene Sequencing Experiments) w/TSM 6.2. HPC infrastructure: RHEL linux clusters using Panasus PanFS iSCSI disk array via 10Gb ethernet To best integrate TSM into their infrastructure, the HPC group wants to backup to iSCSI disk/tape storage via 10GE w/ TSM v6.2 running on RHEL. Our current TSM experience is w/ AIX6.1, TSMv5.5.4.2, FC disk/tape, so we're really trying to "shoot the moon" w/ this new HPC TSM config. !But, hey! It's all going to be new, so why not ...totally different? Please advise, if you have relevant experience/facts, (or theories/opinions, as you believe appropriate;-) FYI, here's how we hope this can work: Because TSM does not support Panasus PanFS, we will use tar to copy completed GeneSeqExprmnts onto (iSCSI) EXT3 FS space @ RHEL linux hosting both TSM 6.2 server and client. TSM client will backup GSE tarballs from EXT3 FS w/ dsmc incremental via ShrMem to TSM 6.2 server, directly onto (iSCSI) LTO5 tape w/compression. Other PanFS volumes will backup via dsmc incremental from NFS mounts @ TSM client via ShrMem to (iSCSI) sequential file STGpool space w/ dedup enabled. (We also hope to use TSM client dedup effectively!-) Thanks for any advice! -- [email protected] (w#203.432.6693, c#203.494.9201, h#203.387.3030) U.S. BANCORP made the following annotations --------------------------------------------------------------------- Electronic Privacy Notice. This e-mail, and any attachments, contains information that is, or may be, covered by electronic communications privacy laws, and is also confidential and proprietary in nature. If you are not the intended recipient, please be advised that you are legally prohibited from retaining, using, copying, distributing, or otherwise disclosing this information in any manner. Instead, please reply to the sender that you have received this communication in error, and then immediately delete it. Thank you in advance for your cooperation. ---------------------------------------------------------------------
