[Beowulf] Mutiple IB networks in one cluster

[Prentice Bisbal]

On 02/01/2014 02:51 PM, Andrew Holway wrote:
> On 30 January 2014 16:33, Prentice Bisbal <prentice.bisbal at rutgers.edu> wrote:
>> Beowulfers,
>>
>> I was talking to a colleague the other day about cluster architecture and
>> big data, and this colleague was thinking that it would be good to have two
>> separate FDR IB clusters within a single cluster: one for message-passing,
>> and the other purely for data movement. I'm a bit skeptical of this myself.
>> I was always under the impression that IB has more than enough bandwidth for
>> message-passing and I/O. I have some questions about this idea:
>>
>> 1. Does this make sense?
> The only project that I have been involved in where we thought about
> this kind of architecture was for Quantum Chromo Dynamics. If your
> application is heavily interconnect and IO bound then there might be
> some value in it however these kinds of applications are somewhat
> rare.
>
> ConnectX 3 has excellent QoS which would probably be a more
> cost-effective way of engineering it.
>
> What is your application?
There is no specific application. This is for a university-wide cluster 
that will be used by many different researchers in many different fields 
using many different applications.

Prentice