[j-nsp] MPC3E oversubscribe rate with two 10x10GE MICs
Tom Storey
tom at snnap.net
Sat Dec 6 18:57:55 EST 2014
As I understand it, it is still split 50/50% wise. Two active MIC slots
arbitrate equally for packet processing resources, regardless of how much
bandwidth they could potentially (or not) use. A 20x1G MIC will essentially
"waste" 40G of throughput. Yeah. :-/
On 5 December 2014 at 19:10, Olivier Benghozi <olivier.benghozi at wifirst.fr>
wrote:
> If you use one 10x10GE MIC and one 20x1GE, on the paper 120 Gb/s would
> mean no oversubscribing, but how the capacity will be really divided?
>
> > Tom Storey <tom at snnap.net> wrote :
> >
> > As was explained to me a while back, the MPC3E has ~120gbit of capacity.
> >
> > But the devil was in how that capcity is shared between the two MIC
> slots.
> >
> > When you have two active MICs that capacity is divided equally between
> the
> > two MICs: 50/50% or ~60/60gbps. It is NOT a case of operate one card at
> > full whack and only use a couple of ports on the other.
> >
> > If you plan to put in a second 10x10 MIC then you'll have to shuffle your
> > circuits around to balance them across the two MICs too.
> >
> > And if you need all wire speed ports then you need the 16x10G MPC3, and
> > only use 3 ports of each PIC.
> >
> > On 30 November 2014 at 23:22, Robert Hass <robhass at gmail.com> wrote:
> >
> >> Hi
> >> I'm currently using MPC3E with one 10x10GE MICs in my MX480 and MX960
> >> routers.
> >>
> >> I need to add 10GE ports, if I will put second 10x10GE MIC in existing
> >> MPC3E what will be oversubscribe rate ? I'm not sure but docs says about
> >> 200Gbps for MPC3E then It should be wire-speed if docs claims
> full-duplex
> >> or 1:2 if docs claims half-duplex.
> >>
> >> What is best solution (from price point of view) to have 16 x 10GE in 1
> >> slot on MX480/MX960 ? MPC3E + 10x10GE MICs or something different ?
>
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
More information about the juniper-nsp
mailing list