BGP-4 Protocol Analysis

Interdomain quality of service routing: setting the grounds for the way ahead

“…The main factors that impact memory requirements at legacy BGP speakers, include the number of IP prefixes/networks (N), the mean AS distance in terms of hop count (M), the total number of unique AS paths (A), the mean number of BGP peers per BGP speaker (S), and the lengths of the binary words required to store a network (R) and to store an AS number (P). Then, an estimate of memory requirement (MR) at legacy BGP speakers is given by MR = (((N * R) + (M * A) * P) * S) [43]. However, when adding a QoS attribute to BGP (e.g., QoS_NLRI attribute) an estimate of MR (ignoring implementation details) at QoS-enabled BGP (q-BGP) speakers can be given by MR q = ((((N * (R + L)) + (M * A) * P) * S)) * C. In the previous equation, L represents the minimum length of the binary word to store the QoS information carried on the QoS attribute and C represents the number of supported services, including the best-effort service.…”

“…However, when adding a QoS attribute to BGP (e.g., QoS_NLRI attribute) an estimate of MR (ignoring implementation details) at QoS-enabled BGP (q-BGP) speakers can be given by MR q = ((((N * (R + L)) + (M * A) * P) * S)) * C. In the previous equation, L represents the minimum length of the binary word to store the QoS information carried on the QoS attribute and C represents the number of supported services, including the best-effort service. Figure 8 illustrates the growth of q-BGP MR for the cases presented in [43], considering L = 4 bytes and up to eight services. As we can observe, it is clear that even for just two or four additional services QoS extensions will undoubtedly require tighter memory requirements to store the additional amounts of routing state information.…”

Experience with the BGP-4 Protocol

“…As discussed in [RFC4274], the driving force in CPU and bandwidth utilization is the dynamic nature of routing in the Internet. As the Internet has grown, the frequency of route changes per second has increased.…”

“…In this way, BGP is work conserving [RFC4274]. In cases of extremely high route change, a higher volume of route change is sent to those peers that are able to process it more quickly; a lower volume of route change is sent to those peers that are not able to process the changes as quickly.…”

Inter-domain routing for tactical mobile ad-hoc networks

“…The BGP Finite State M achine (FSM ), as described by RFC 4274[15], is a set o f rules that is applied to BGP peers relationship. The FSM m ust be initiated and maintained for each new incoming and outgoing peer connection.…”