[Info-vax] PEA0: Errors

[John Santos]

In article <idqdnVgINPYaYjPUnZ2dnUVZ_ryWnZ2d at giganews.com>, rgilbert88
@comcast.net says...> 
> Christoph Gartmann wrote:
> > Hello,
> > 
> > we have errors on all PEA0: device in the cluster. These errors increase from
> > time to time, that is, the numbers are stable for a few hours or days and then
> > they increase rapidly during a minute or so. How do I find out what is going
> > on? The switch that connects the cluster is a HP Procurve 3500yl.
> > 
> > Regards,
> >    Christoph Gartmann
> > 
> 
> How many nodes in the cluster?  I don't recall ever seeing a cluster 
> with a switched interconnect.  I have run a two node cluster with the 
> two nodes connected back-to-back with a crossover Ethernet cable.  I 
> have seen the same thing done with a hub.  I see no reason why a larger 
> cluster could not use an Ethernet (UTP) hub.

In my experience, an Ethernet switch is infinitely superior to a hub for
an NI cluster.  We no longer get swarms of errors and get *much* better
performance with switches.  Plus the switch copes much better than a hub
with old, slow (10Mb HD) Vaxes and Alphastations, while letting the 
newer equipment talk at full speed.


-- 
John Santos
Evans Griffiths & Hart, Inc.