looking for SLA examples

We are looking at putting in contract language surrounding Service Level
Agreements and I would like to know what people are doing from the
standpoint of latency, packet loss, and bandwidth delivery capability.
What examples do people have written in contracts on the above, and what
penalties are there for not resolving problems?

    Service Provider agrees to provide maximum latency guarantees
    1) on-net CPE to on-net CPE 60 msec.
    2) on-net CPE to off-net peering point 30 msec.
    3) End to End 100 msec within the continental US.
    Remedies for violation of latency guarantee for unresolved latency
    1 month 15% off of monthly rate
    2 month 25% off of monthly rate
    3 month 35% off of monthly rate
    >3 month customer has option of cancelling remainder of contract.

My understanding is that expected latency should be based upon the
speed of light in glass 0.6C or roughly 6msec per 1000 fiber route
miles plus some amount for router serialization and queuing delays
(~5 msec/hop?). What do people use as standard rule of thumb for
calculating latency? What do people expect for service level
commitments, normal+50% (100?) ?

Tim Peiffer peiffer@nts.umn.edu
Operations Manager - Engineering Automation
Networking and Telecommunications Operations
University of Minnesota +1 612 626 7884 desk
2218 University Ave +1 612 625 0006 problems
                 +1 612 626 1002 fax
Minneapolis MN 55455, USA

5ms per hop? Assuming >=OC-12 backbone, that's way
excessive. At OC-12, a 1500 byte packet takes
20 microseconds to serialize. OK you will statistically
encounter a few queuing delays even on a near-idle
backbone (think about number of other
ingress interfaces), but again these should
be a few packets when the circuit is not
loaded. So 0.5ms per hop seems generous.

Remember that with cloak-of-indeterminacy technologies
such as ATM and MPLS, you may not be able to see many
of the hops.