Hi Pavel,
HQoS is needed for subscriber aggregation, exactly for the case that is
discussed in another thread "Juniper buffer float".
NG cards without HQoS can be configured for "flexible mode" which is
HQoS limited to 32k queues. It's suitable for example if you have some
number of vlans and want to do per-vlan shaping and queueing. For
example, 32k queues with 4 queue per interface will allow to serve 8000
subinterfaces. On MPC2E-NG it means that you can do per-vlan QoS on 8
10G ports with 1000 vlans on each port - not bad in compare with old
non-NG non-Q cards that can do queueing only on physical interfaces. It
will work for small-scale LNS, up to 8k subscribers, but watch out to
not exceed it ;)
Full-scale NG HQoS card allows 512k queues that allows (in theory) to
terminate 64k subscribers with 8 queues per interface.
Kind regards,
Andrey
Pavel Lunin писал 20.10.2018 05:12:
Hi,
If memory serves, MPC2-NG is much like MPC3-NG: one MPC5-like PFE
inside. A
good question is what is the difference between the MPC2-NG and
MPC3-NG...
I'll let the astute readers figure it out on their own.
Old MPC2 non-NG is the "first generation Trio"-based card: two PFE,
one per
MIC. There is no much sens in buying it today, it's a 10-years old
thing.
E vs. non-E - Saku is (as always) right, the non-E was the very very
first
generation of MPC cards, which had a kind of broken oscillator (only
matters for SyncE applications). So we normally omit the E in every
day
language today, as all MPC cards are E since many many years.
And yeah, "HQoS" (aka -Q/EQ version of cards) has nothing to do with
the
NG/non-NG story. Most cards have a -Q version to support "HQoS".
Shortly
speaking, it's for those folks who don't know what to do with their
employer's money.
--
Pavel
_______________________________________________
juniper-nsp mailing list [email protected]
https://puck.nether.net/mailman/listinfo/juniper-nsp
