The paper [LKH02] by Hyo-Jin Lee, Korea mentions a mapping mechanism between QoS parameters (here, it means application level QoS parameters) and Network Performance Metrics (NPM). Here, these people have given an architecture of a SLA monitoring system and have given a method to map QoS to NPM. They are experimenting with their mapping function by applying it to IP backbone network.
A SLA usually mentions the following QoS parameters: MTBF, MTRS, Delay, latency, delivery and availability. Now, given above parameters we should be able to map them to network performance metrics. We should also be able to influence and use this information of SLA breaking for billing, reporting and even, use as a feedback for the network to restore its status such that the network again conforms to the SLA. This kind of system is known as Service Level Management System. It performs 2 functions: SLA monitoring and SLA provisioning The architecture of the same is given in the paper mentioned above. Here, we collect NPMs from the network through SLA agent and then map them to QoS. Say, some application wants a delay of less than 10 sec. We should be able to collect the Network state (which is nothing ,but NPMs which are: traffic congestion, throughput, availability and latency).
Above mapping as done in paper is shown for network access services like IP-VPN, leased-line service, xDSL. It is yet to be extended to application and content services like web-hosting etc. But, this scheme is quite generic and is not precise. It can only serve as a guideline to serve a particular need or a particular situation wherein the entire mapping mechanism would need to be reworked.