[j-nsp] Stackable switches, looping stacking ports

[Tom Storey]

So I imagine that might help with latency, but is it going to have any
affect on bit rate throughput?


On 9 April 2013 21:05, Chuck Anderson <cra at wpi.edu> wrote:

> On Tue, Apr 09, 2013 at 11:48:36AM -0700, joel jaeggli wrote:
> > On 4/9/13 11:15 AM, Tom Storey wrote:
> > >Hey all.
> > >
> > >A colleague of mine tells me that, if you have a single stackable switch
> > >(not in a stack obviously) and do not loop the two stacking ports on the
> > >back using the stacking cable that comes in the box, then you reduce the
> > >effective throughput of the switch.
> >
> > The ex4200's asic has a capacity of 136Gb/s from the front panel
> > ports which is 100% of line rate across all ports. I don't imagine
> > connecting the asic back to itself is that useful or usable
> > topology.
>
> It does make a positive difference to loop back the stack cable on a
> standalone unit.  See this diagram:
>
> http://blog.cochard.me/2010/08/juniper-ex-4200-internal-pfe-routing-in.html
>
> Connecting from ports 0-23 to 24-47 has to normally cross 3 internal
> PFEs.  If you connect VCP-0 to VCP-1, then it only has to cross 2
> PFEs.
>