Re: VMS Shadowing vs RAID - your opinions wanted

In <sedwards.795773013_at_crash.cts.com>, sedwards_at_crash.cts.com (Newline) writes:
>I'm interested in your opinions on the relative merits of VMS shadowing
>vs various RAID levels.
>
>Some of the issues that come quickly to mind are: Cost, Performance,
>Reliability, Expandability, and Processor loading.
>
VMS Host Based Shadowing is primarily a data availability product. It is intended to to allow you to survive the failure of various components in the i/o chain.

For a mostly-read i/o mix it does provide a performance boost by fanning out reads amongst the various shadow set members. For many configurations it will also respond to hardware failures faster than the existing MSCP-based failover mechanisms (often by expelling the misbehaving shadow set member, though it can be added back to the shadow set later).

Host load is mainly the expense of issuing writes 'n' times to an 'n' member shadow set. It will attempt to use the least costly path for reads. In an ethernet or FDDI config you may see the cost of issuing writes multiple times across the wire.

I can't really comment on RAID alternatives without knowing what you are considering (note that VMS shadowing is of course RAID-1). If you are looking at hardware RAID controllers, you gain performance, sometimes at the expense of availability. You can of course shadow hardware RAID devices, but the optimum is to build RAID arrays out of shadowed member units.

Gary Hughes (ex-VMS Cluster developer) Received on Wed Mar 22 1995 - 04:23:46 CET