Proc. SPIE. 6022, Network Architectures, Management, and Applications III
KEYWORDS: Mirrors, Switching, Weapons of mass destruction, Networks, Computer simulations, Local area networks, Line edge roughness, Electronics engineering, Forward error correction, Information science
MPLS forwards the labeled packets on the Label Switching Paths (LSPs) which are built between ingress/egress Label Edge Router (LER) peers. As MPLS is the mirror of the IP layer, supporting the reachable capability to <i>n</i> destinations means that the MPLS network needs to build <i>O</i>(<i>n</i>) LSPs (<i>n</i> is determined by the granularities of the labeled packets). As the result, if it needs to set up a full mesh connection between each ingress/egress LER peer, <i>O</i>(<i>n2</i>) LSPs are required. When the size of the MPLS network increases, the cost for maintaining the LSPs will severely increase. It is a serious problem for the scalability of MPLS networks. In order to achieve the scalability of MPLS networks, a network aggregation scheme based on weighted LER dominating set is proposed. The distributed construction method of the LER dominating set is introduced. The aggregation scheme can reduce the amount of the egress nodes when establishing LSPs so as to form an aggregated virtual backbone topology. At the same time, considering the bandwidth between the dominator and its dominatees, a bandwidth-constraint weight is used when constructing the dominating set. As the result, the dominating set has the optimal bandwidth feature and this scheme can apply to the network aggregation with other single constraint.
This paper analyzes a communication network facing users with a continuous distribution of delay cost per unit time. Priority queueing is often used as a way to provide differential services for users with different delay sensitivities. Delay is a key dimension of network service quality, so priority is a valuable resource which is limited and should to be optimally allocated. We investigate the allocation of priority in queues via a simple bidding mechanism. In our mechanism, arriving users can decide not to enter the network at all or submit an announced delay sensitive value. User entering the network obtains priority over all users who make lower bids, and is charged by a payment function which is designed following an exclusion compensation principle. The payment function is proved to be incentive compatible, so the equilibrium bidding behavior leads to the implementation of “cμ-rule”. Social warfare or revenue maximizing by appropriately setting the reserve payment is also analyzed.
Multi-protocol label switching (MPLS) is multiprotocols both at layer 2 and layer 3. It is suggested to overcome the shortcomings of performing complex longest prefix matching in layer 3 routing by using short, fixed length labels. The MPLS community has put more effort into the label switching of unicast IP traffic, but less in the MPLS multicast mechanism. The reasons are the higher label consumption, the dynamical mapping of L3 multicast tree to L2 LSPs and the 20-bit shim header which is much fewer than the IPv4 IP header. On the other hand, heterogeneity of node capability degrades total performance of a multicast group. In order to achieve the scalability as well as the heterogeneity in MPLS networks, a novel scheme of MPLS-based Multi-area Layered Multicast Scheme (MALM) is proposed. Unlike the existing schemes which focus on aggregating the multicast stream, we construct the multicast tree based on the virtual topology aggregation. The MPLS area is divided into different sub-areas to form the hierarchical virtual topology and the multicast group is reconstructed into multiple layers according to the node capability. At the same time, the label stack is used to save the label space. For stability of the MALM protocol, a multi-layer protection scheme is also discussed. The experiment results show that the proposed scheme saves label space and decrease the Multicast Forwarding Table in much degree.