ABSTRACT
To have an effective end-to-end QoS capability in the Internet, it’s necessary for the Border Gateway Protocol (BGP) to provide QoS enforcement among the inter-BGPs as well as IGPs. Current BGP4 operates based on the policy, and policies are represented by attributes such as LOCAL_PREF and MED to influence incoming/outgoing traffic flows on purpose. These attributes are optional nontransitive or well known discretionary. So these attributes can influence the preferences between directly-connected neighbor BGPs, but are not sufficient to provide general end-to-end Traffic Engineering. This paper proposes an extended BGP routing algorithm which can provide an end-to-end Traffic Engineering capability. This new method doesn’t require additional BGP message type, nor change a structure of existing routing tables. It only uses a slightly changed operation of AS_PATH attribute, which is well-known mandatory. Simulation shows that the new BGP algorithm increases link utilization among ASs and can be used as compulsive load balancing mechanism.
