Case: Storage and clustering with shared storage

* Below is my original inquiry, I corrected the part
where I made a mistake, which confused Bob Koehler. *

This is something I've been wondering about for some time, on the topic
of storage and clustering with shared storage. I haven't been able to
find much in documentation, or not specific to my question, so I'll
hereby be inquiring.

I have a U320 SCSI RAID disk array, offering two physical channels (as
is most common). I added an extra dual channel U320 HBA, an HP A7173A
(LSI1030) in one of my systems, so there should be four buses total
(PKA, PKB, PKC and PKD).

The way I have things, or prefer to have it, is that PKA and PKB will
be for internal disks and with PKC and PKD are available for the RAID
disk array. There's one RAID level 5 logical disk, split up into 7
'slices' (read: volumes mapped to "LUNs") of <=2 TB (since that's what
appears to be the maximum that VMS will be able to address, since V8.4).

An import detail: The disk array is currently hooked up to one node,
which is clustered via VMScluster, with MSCP, TMSCP and DECnet set up
over 10 GbE. 10 GbE works great, its performance is great. Besides
redundancy, what advantage would there be? (If it was over a FE or
GbE network, then I'd see a performance benefit. Though, if I'm see-
ing this incorrectly, please let me know.)

Besides the clustering factor. On a single host, with two physical
SCSI channels at one's disposal, is there a way or method to configure
the two disk array channels in an 'arbitrated loop' like setup? (As
seen with FC SAN setups). So, that in theory one would be able to
increase throughput to ~6 gigabit (~640 megabyte/second; though more
likely with RAID level 0).

Any thoughts, advice and so forth will be appreciated.

- MG
.

Relevant Pages