The Question is:
I am responding to a tender that will replace three AS8400's that are clustered
running OpenVMS V7.1 ( two machines in one location and another 20 KM away
connected via Gigaswitches and FDDI ). Storage is HSJ based and three way
mirroring via volume shado
wing is used.
New Alphas, two per site and Fibre Channel Storage at each site is being
proposed to replace the current equipment.
The current DataVault FDDI service used for Cluster/LAN traffic is being phased
out and Gigabit ethernet is being proposed as its available from a local
Communications provider ( ATM at this point is too expensive ).
My questions are :-
1) What is the max distanced Gigabit ethernet clusters can be run over, I
beleive its 240KM is this true, are there any obvious restrictions not
mentioned in the clustering manuals.
2) Jumbo frame sizes are supported in V7.3 ( ~7000Bytes ) will this improve
cluster traffic speeds over a standard Gigabit Ethernet shared service from a
local Telco. What are the other advantages.
2) The service provider has quoted around a 2ms delay over the 20Km link
between sites- is this OK for Cluster traffic between the two sites assuming
the gigabit switches at either end will support autonegotiation.
3) Volume shadowing will need to run between sites I'm assuming that because
mount verify timeout values start at 1 second that 2ms delays will not be an
issue for any minimerge / copy operations ( around 20 shadow sets planned )??
are there any potential
4) Are there any reference sites running multisite clustering over Gigabit
Ethernet via Telco connections, are detailed configurations available ? - who
are the customers ?
5) Are there any Specific whitepapers etc on Gigabit Ethernet Multisite
Clusters with Fibre Channel Storage. ( Not interested in CI DSSI SCSI etc )
Any help on the Above questions would be greatly appreciated.
The Answer is :
1: Please see the OpenVMS Cluster Software Product Description (SPD)
for the current maximum supported diameter. Between the SPD and
the manuals, the OpenVMS Wizard expects the "obvious restrictions"
will be listed.
2: Unclear, whether or not the larger frames are useful depends on
3: The Mount Verify Timeout setting depends on the site-specific
trade-offs between requirements for ride-over vs rapid fail-over.
4: Yes, but you already knew that you were not likely to receive a
pointer to a reference site via the Ask The Wizard area.
5: A cluster is a cluster is a cluster. Read the existing cluster
manual. If you wish to know about Fibre Channel requirements,
read the Fibre Channel manual referenced in the FAQ.