Byung Jun Ahn

Performance evaluation of the flow-based router using intel IXP2800 network processors

Exponential increment of users and emergence of realtime services with QoS requirements cause a great change of Internet traffic characteristics from early best-effort characteristics. The currently developed packet-based router with MPLS and DiffServ technologies has still limitations on providing services with QoS guarantee. Hence, a flow-based router is introduced for dealing with realtime services with QoS requirements. This paper proposes a flow-based router using Intel IXP2800 network processors. By imposing flow-awareness into IXP2800, the proposed flow-based router has basic functionalities such as flow classification, flow maintenance, flow-based forwarding, and flow-based congestion resolution. And we analyze performance of the proposed flow-based router considering flow characteristics. The flow maintenance duration gives a great effect on its performance. Compared with the packet-based router using IXP2800 network processors, performance of the flow-based router is enhanced in terms of the packet processing time.

Hybrid Medium Access Control Scheme to Enhance the Performance of IEEE 802.11s Wireless Mesh Networks

In recent years IEEE 802.11 has become the most famous protocol for implementation of wireless local area networks (WLAN). The diffusion and the increasing demand for WLAN coverage make a lot of installation of access points (APs). Although the cost of APs is lower than the cost of cellular equipment, interconnecting APs through wired links requires high cost and complexity. Therefore wireless mesh networks (WMNs) that directly interconnects APs through wireless links become a promising technology for the next-generation wireless networks. One of the important functions in WMNs is multi-hop packet forwarding between source and destination. However, in WMNs based on the IEEE 802.11 DCF, there are many collisions and retransmissions in the nearby mesh portal (MPP) which is bottleneck node of WMNs. In this paper, we propose hybrid medium access control (HMAC) scheme for WMNs and compare performance of the HMAC with that of conventional DCF in terms of throughput and delay. From the result, the HMAC yields better performance than the conventional DCF for WMNs.

Performance Analysis of an IP Lookup Algorithm for High Speed Router Systems

IP lookup is an important issue in designing high speed router systems. The most important function of router is to determine the output port of an incoming IP packet according to its destination address. In this paper, we evaluate the performance of the IP lookup algorithm in IQ2200 Chipset for high speed routers in terms of the memory required for storing lookup information and the number of memory access on constructing the forwarding information and then propose a scheme to enhance the IP lookup performance.

Improved VC-merging method in MPLS networks

VC-merging allows VCs to be mapped onto the same VC label. In the multiprotocol label switching (MPLS) domain, an ATM-LSR (label switch router) needs a VC-merging method because of network scalability problems. VC-merging also plays an important role in providing multipoint-to-multipoint multicast and share-tree multicasting paradigms. When VC-merging is applied to multiple streams of incoming packets, all the VC-merged cells have the same label at the output port of the ingress node. If those VC-merged cells are transmitted using AAL5, cell interleaving occurs. Thus, at the egress node of VC-merging, an identification process is required in order to distinguish cells belonging to a packet from other cells. Various approaches have been proposed to help this identification process. In this paper, we propose improvements to the traditional VC technique to minimize the need for additional buffers at intermediate merge points to decrease end-to-end delay. This improved VC-merging mechanism requires receiving nodes to have a queue corresponding to each queue of the sender node, such that cells from a particular queue of the sender node are queued in a corresponding queue of the receiving node. In order to distinguish VC-merged cells from non VC-merged cells, an information cell is used. Upon the arrival of a cell at the sender node, the scheduler of the sender node transmits the cell immediately based on a round-robin algorithm. Simple equations in this paper show that the improved VC-merging method has advantages against the traditional methods in terms of buffer size and end-to-end delay time. Furthermore, if we change the scheme of queue searching (e.g. weighted fair queueing), QoS can also be supported.