RE: [j-nsp] Cisco equivilant to an M5?

From: Guy Davies (Guy.Davies@telindus.co.uk)
Date: Fri Dec 07 2001 - 04:25:30 EST


 
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

> * andyh@verio.net (Andy Harding) [Thu 06 Dec 2001, 23:15 CET]:
> >> The M5 can be stuffed with 48 of them you'll still be able to
> >> run them all at full capacity.
> > this card over-subscribes the backplane - juniper themselves
> > readily admit this
> >
> > you can't run all these ports full-rate
>
> Good point. I should've done the math myself, too...
>
> Are they making a two-port GigE PIC yet? Those are handy in
> redundant setups, where you use either one of them at any time.
> Doesn't seem like
> it - my employer at that time going bankrupt probably didn't
> add to the
> credibility of the request. :-)

No. The M160 has 2 port and 4 port GE PIC but all other units have
only got 1 port GE PICs. There is a 4 port quad wide PIC for the
M5/M10/M20/M40 which basically fills a complete slot with GE and is
inherently oversubscribed (3.2Gbps to the backplane) but between
interfaces on an M5 you should get full line rate because there is no
way the packets have to go over the "backplane". All data stays in
the on card memory of a single FPC. The only limitation is whether
the FEB has enough packet memory to sustain 4 Gbps bidirectional
traffic. In the M10, even though the ingress and egress ports may be
on the same FPC (or even the same PIC), the data may be written to
the other FPC, thus having to cross the backplane (twice) to make its
journey from ingress to egress.

The M5 is a bit pricey for a four port GE router (even if they can
all do full line rate). The card becomes more useful on the M10/M20
where you can have four oversubscribed GE going down into a
datacenter, for example, and having other ports there for the uplinks
(GE/OC12/OC48).

Regards,

Guy

-----BEGIN PGP SIGNATURE-----
Version: PGP 7.0.1

iQA/AwUBPBCK843dwu/Ss2PCEQIkvQCfaikTpfoAN4EN80EVeWfC9pVwN0MAoLki
QgEly+359KRRPmQdjwpKxrS/
=XEPs
-----END PGP SIGNATURE-----

.



This archive was generated by hypermail 2b29 : Mon Aug 05 2002 - 10:42:38 EDT