Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |  

This web mail archive is frozen.

This page is part of a frozen web archive of this mailing list.

You can still navigate around this archive, but know that no new mails have been added to it since July of 2016.

Click here to be taken to the new web archives of this list; it includes all the mails that are in this frozen archive plus all new mails that have been sent to the list since it was migrated to the new archives.

Subject: Re: [OMPI users] [mvapich-discuss] MPI-2 over ibverbs/rdma
From: Vivek Satpute (vivekonline86_at_[hidden])
Date: 2009-11-27 02:05:02


On Thu, Nov 26, 2009 at 10:35 PM, Chuck Ritter <cfr100_at_[hidden]> wrote:

> On Thu, Nov 26, 2009 at 6:30 AM, Vivek Satpute <vivekonline86_at_[hidden]>
> wrote:
> > Hi Ritter,
> >
> >> > And why those examples in libibverbs-1.1.2 package takes IP address ?
> >> > Do you have any idea about that ?
> >>
> >> Why not?
> >>
> >> I think hostnames and ips were used because they uniquely identify a
> >> node. Why create a parallel naming scheme? It just adds more data to
> >> track.
> >>
> > I agree with this that why there should be parallel naming scheme. But
> > Infiniband devices already have LID and GID naming system to identify
> > devices
> > So instead of using IP address to uniquely identify, we can use those
> > LID/GID.
> > Am I right ?
> That would have been another possibility.
> You could say the same about the SCSI bus. LUN and bus information is
> mapped to an abstract device so that the user does not have to deal
> with those low level details. In the case of SCSI, the OS handles
> that... maps it into devices. I think IB somewhat follows that model.
> Yope. Maybe you are right.

> C