Quality of Service
Quality of Service |
|
It becomes difficult to guarantee QoS when data crosses a WAN. When traffic is carried end-to-end, for example, between remote offices, it traverses multiple domains, including the Internet. At the boundary of each domain, traffic is passed from one service provider to the other to cross the Internet. The different providers must negotiate agreements on how they must carry and handle each other's traffic in order to ensure QoS. These negotiated contracts are called service level agreements (SLAs).
SLAs specify the rate at which traffic from one provider (the customer) must be carried by the other provider–usually the ISP. The administrator for each domain must ensure adequate network resources are available to support SLAs offered by their domain. The SLA might specify classes and marking rules. In addition, when providers agree to carry a customer's traffic, they are agreeing to consume resources in their own network that might otherwise be sold to another customer. For example, some traffic flows consume more resources, making that flow expensive to carry. As a result, most providers limit the amount of traffic that they agree to designate as priority. This limitation is referred to as policing, and these limits are negotiated in the SLA. If packets arrive that exceed the agreed-upon SLA, the provider discards the flow or demotes the priority of the flow to one that meets the terms of the SLA.
In general, the source domain marks the priority of the packets before they leave the source network boundary, because it is easier to determine the appropriate marking closer to the sending application before the packets have been aggregated with other flows by another provider. However, sometimes another provider is requested to mark packets on behalf of the customer, such as when the customer is a legacy network. In such a case, the provider marks the packets only to the extent that the priority rate specified does not exceed what has been negotiated in the SLA.