Showing results for 
Search instead for 
Do you mean 

Tru64 Cluster

Occasional Visitor

Tru64 Cluster

We have two Alpha ES40 servers clustered together, server named A & B . These servers are located in the same place.

We have an another server named C (alpha DS 20)located 15 miles away from A & B servers. Is it possible to cluster server C with A & B. Our OS is Tru 64 5.1 A . This question arise because the server is located 15 miles away.
5 REPLIES
Honored Contributor

Re: Tru64 Cluster

Hi,

for a cluster you need shared disks and a fast cluster interconnect. I wonder, how you are going to provide this over this distance.

Michael
Honored Contributor

Re: Tru64 Cluster

15 miles is too far, but the right configuration seems to be optical gigabit interconnects and a fibre channel storage connection. I'm not sure if SAN networks are certified for that distances...

Optical memory channel interconnect are limited to 1-2 miles.


Help() { FirstReadManual(urgently); Go_to_it;; }
Respected Contributor

Re: Tru64 Cluster

Hi, you can cluster at least 2 nodes together using ethernet now instead of memory channel interconnect. Would this mean that you could achieve clustering over 15 mile radius?? I might be easier if 3 node was part of a private WAN.

We have a similar private WAN with ISDN 128k burst in the UK with machines located 200 miles apart, and we seem to encounter no problems. Maybe you could acheive what you are after if you are able to implement a WAN
Trusted Contributor

Re: Tru64 Cluster

Be carefull what you build, it may work. But be sure to check that it is also supported by HP.
Watch, Think and Tinker.
Honored Contributor

Re: Tru64 Cluster

There are some configuration rules described within the cluster installation guide.

The speed of the interconnect is important, a wan connection will not be fast enough, that is the reason using optical cable stretching the memory channel connections up to 2km (but with slower response for the whole cluster). If the cluster will be booted and the interconnect requests are not answered within a time period the whole cluster will initiate a crash....

In case of a mission critical system the "supported by HP" statement is very important.

Help() { FirstReadManual(urgently); Go_to_it;; }
//Add this to "OnDomLoad" event