asymmetric SCA round trip time estimate

[David Jones_21]

I have a small cluster (OpenVMS 8.3), DS15 boot server and DS10 satellite. I'm trying to get a better feel for how the ethernet connections can best be used for the cluster traffic. I set the priorities so the preferred channel was between EIB (DS15) and EWB (DS15), now show scacp show channel reports a fairly consistent RTT on the DS10 (2000 - 4000 RTT) but a widely varying RTT (4000 - 17000 usec) on the DS15 that is 2 times or more the DS10's value. Is to be expected, given one machine is slower and tends to be on the receiving end of MSCP transfers?

P.S. After the fiddling with stopping/starting devices in SCACP, the DS10 no longers 'sees' EIA on the DS15 at all, though the DS15 still seems aware that the remote DS10 has both 1 EWA and EWB.

[John Gillings]

David,

I'd suggest starting by just cross connecting everything and letting the systems work it out for themselves.

If you don't have a noticeable problem, don't try to configure anything. The SCA traffic should dodge around according to whichever path looks best at the time.

I suspect that most of the time decisions made to block traffic from a particular path are not made for technically justified or even justifiable reasons. Leave the cluster maximum flexibility to do whatever it thinks is right, and reduce your management workload.

(let's face it, given the members of your cluster, it can't be too performance sensitive... if you need to "tune" it by adjusting SCA to make it viable, you'd be better off just upgrading the hardware)

[David Jones_21]

Yes, I know it can figure it all out. I want to better understand what the governing factors are.