Internet link qos (looks a bit complicated)

[laszlo.pal]

Currently I'm facing with a task where I need some advise from someone who's more experienced in Qos configurations. Do we have some forum or other resource where I can ask those questions?

The situation as the follows:

We have a customer where we manage Internet router and internet access. Currently we have two internet link, one for "normal" users other for VIP users. This is not the perfect configuration and not fully working as we expect, so I think it would be better if put together the two links and implement some kind of traffic prioritization on the links.

The expected behaviour should be this:

- "VIP" type customers should have "reserved' bandwidth, or high priority over another traffic. They can be classified by source address

- One of the traffic type should have same priority as VIP users (this is scp traffic). This traffic can be classified by destination address and destination port

- Every other user should have lower priority and some kind of 'per session maximum rate' should be implemented to avoid circuit fill-up by one user

Do you think it is possible?

Thank you

Pál, László

Cisco CCNA

HP AIS - HP OpenView Network Services

[spremkumar]

hi

QOS in internet you cant predict the results as you desire to have.

Untill your end you can preserve the qos settings and reserver the kinda b/w for the privileged customer but once its out from your network/end the qos configs/values wont be referred since QOS mostly being deployed in SP backbone to gurantee the kinda service opted by the customer between their locations.(eg.customers having multilocation hybrid connectivity connected to the same SP in all his location).In this kinda scenario the SP has its own control over the whole backbone from source to destination and he can have similar kinda configs and treat the traffic in similar way all throught his network.

Coz QOS requires the settings/configs to be the same from the starting point to the end point otherwise it may not give the desired results.

in your case if you have some 2 or 3 boxes in transit before reaching the upstream or internet cloud and if you really want to have some kinda mechanism to improve the latency within your network i would suggest to give a try like this ..

you can create class-maps for the customers matching the ACLS allowing the traffic being propagated from the customer network .

mark it with EF bit or AF31 based on the customer type and requirement ,apply it on the in direction on the interface in which the customer is getting connected.

create another class-map matching ef and under another policy map u can give priority b/w for the same class of traffic.

same kinda of uniformity in the configs can be followed and checked out .

i guess the other way around is to limit and drop the excess traffic using CAR(rate-limit) and shaping but thats again depends on the customer needs and the kinda services being offered to the custome..

regds