Re: [Veritas-bu] multipathing with tapes on Linux?
2011-07-27 17:14:51
On Wed, July 27, 2011 17:03, Len Boyle wrote:
> David,
>
> I think the answer depends on what you mean by multi-pathing, and the
> maker of the tape drive.
>
> The new IBM tape drives include two ports. But I think that it is only for
> failover. And I suspect that the support would only be in the IBM tape
> driver. Which was released with it's source code the last I looked.
>
> I also suspect that you do not need multi-pathing to feed the tape drives.
> The native speed for a lto-5 tape drive is greater than that of a gige
> card. And if you send data that can be compressed 2-1 or 3-1 then even
> more so.
>
> The hard part is getting the data off the disk fast enough to drive the
> tape drive.
We have sixteen LTO-4 drives that we want to drive with as few media
servers as possible for support reasons. If we could get a few servers
with 10 GigE and and multiple HBAs (or a single multiport HBA), then we
could (attempt) saturating the 10 GigE and having 8 or 12 Gbps of FC on
the other end going to multiple drives in a library.
If multi-pathing is not supported, another option would be to configure FC
zone such that: drives 0-3 are only visible to HBA0-port0, drives 4-7
visible to HBA0-port1, drives 8-11 visible to HBA1-port0, and drives 12-15
visible to HBA1-port1.
We'd prefer to have "normal" multi-pathing if possible, but spreading the
tapes to different HBA ports is another possibility.
We're not concerned with multi-pathing on the drives, but utilizing as
much bandwidth from the clients to the media server, and from the media
server to the library/drives.
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|
|
|