Reg. LAN configuration...

[Accullise]

I have 4 alpha servers with openVMS7.3-2 having MC and LAN clustering and one Itanium server with openVMS8.3h1 only in LAN, not in MC. So i want to add a new lan cluster for all 5 servers.

1.what are the procedures for gigabit ethernet clustering and what are the parameters wants to consider with that?

and

2.what is the best clustering(ethernet,FDDI,etc.) for my case?

3.what are the drivers priority level (for ex. pa driver(MC) having high preference than pe driver(ethernet LAN)).so can anyone List that?


Regards,
Saravanan.M

[labadie_1]

Hello

2) It seems the best Clustering is Lan, as you need at least the Cluster protocol (SCA) on a network device, if you want to communicate with the other members of the Cluster.

3) The order for Interconnect preference is
(extract from the very good document of Keith Parris, search for f2000_vmscluster_seminar_basic_intermediate.ppt on his website)

Galaxy Shared Memory Cluster Interconnect
Memory Channel
CI and DSSI, by adapter bandwith
FDDI and Ethernet, by
maximu transfer size
lower latency

[Jon Pinkley]

1, gigabit ethernet is ethernet. Configure that same as you would ethernet. Get it working, then worry about tweaking things like jumbo frames.

2. Gigabit ethernet is currently the best of the ones you have listed, and is in general recommended for all new configurations.

3. Don't worry about this until you have it working. The defaults work. Worry about tweaking if you think you have a problem after you get things working.

All your questions are answered in detail in the VMS Cluster documentation, which you can download in PDF format, or browse online in html format.

Jon

[Hoff]

"Best" is not a decision that can be made in isolation, barring those few and rare cases of an unlimited budget.

For typical cases with typical budgets (and what's "typical", he asks?), Gigabit Ethernet (GbE). 10 GbE, if you're really pushing your LAN hard. Shared SCSI and FC SAN and/or multiple parallel LANs, if you have the right gear and right configurations available and the right requirements, and then as an adjunct to GbE or 10GbE to off-load the storage traffic from the LAN.

Here? No FDDI. No Galaxy. No MC. No anything else.

No premature optimizations should be considered here. Get the NICs and switches and related parts installed and working. Then and only then consider optimization, and then only if you need to optimize. And that's usually (and what is "usual", he asks?) optimization around redundancy or around offloading storage I/O.

And as has been commented here, do please take the time to read the cluster manuals.

Rather than looking for network optimizations, I'd first look to get those OpenVMS Alpha systems up to the current V8.3 release, too. This is not a major upgrade for OpenVMS Alpha, but using consistent releases does have an advantage in maintenance and upkeep. (You can use common files and current DCL features without regard to the versions, for instance.)