[Simh] Mixed-node OpenVMS Clustering

Hittner, David T (IT) david.hittner at ngc.com
Tue Mar 10 08:34:52 EDT 2009


VAX and Alpha/Integrity Clustering works just fine in practice.

HP states that VAX and Integrity clustering can be done for migration
purposes. This does not mean that they plan to take away the
functionality - far from it. They plan to maintain cluster compatibility
for as long as possible, because that's the whole point of clustering.
When HP states that they don't warrant VAX-to-Integrity in production,
it really means:

	"HP doesn't have enough real hardware and/or time to accurately
verify and debug all configurations. If you try it and it works for you,
great. If you can't get it to work, we'll try our best to help, but we
are not obligated to make it work."

The SCS Clustering protocol depends on timely and accurate multicast
packet delivery; duplicate, dropped, or delayed packets are the killer
of clustering. Cluster success on any slow LAN/MAN/WAN sometimes
requires tweaking the cluster parameters, particularly the timeouts -
the default cluster parameters are for a "standard" small-to-medium LAN.
SIMH clustering works pretty well, as long as the host doesn't delay the
SCS packets too long - there are ethernet device settings ("poll", etc.)
to increase the performance of the SIMH ethernet behavior, at the
expense of lowered CPU performance of the SIMH VAX.

Having written a lot of the SIMH ethernet code, I made sure that SIMH
VAXes would cluster with real VAXes. After testing, I found that the
SIMH VAX can also cluster with real Alphas. Having just added an
Integrity to my hardware, I fully expect that clustering with the
Integrity will work once I have time to fire the Integrity up. :-)

Dave Hittner



More information about the Simh mailing list