Special issue on "Challenges in high-performance switching and routing in the Future Internet"

High-performance, high-capacity packet switches and routers in the metro and backbone network are a basic infrastructure requirement to support the Future Internet. Trends in traffic evolution point to higher traffic volumes and increasingly unpredictable traffic variations caused by the mobility of large multimedia flows. These tendencies translate into challenges to switching and routing in the aggregation and core networks. This special issue is an effort to put together original papers on the design of high-performance switching and routing systems to address these challenges, including architectural, planning, traffic engineering and algorithmic aspects of electronic and optical high-speed networks. In response to the call for papers for this special issue, we received 43 submissions for consideration for publication. After a rigorous two-step review process, we accepted 12 high quality papers reporting state of the art in highperformance switching and routing. Seven out of the twelve papers included in this special issue [1–7], report novel and surveying proposals addressing the wire-speed electronic packet-by-packet processing at multi-Gbps. As traffic grows, nodes are required to aggregate thousands of user packet flows, while meeting their QoS requirements. This entails boards designed with dedicated architectures (ASIC or FPGA based) or specialized programmable devices (network processors). The packet processing cycle implemented in the line cards includes (i) packet classification to identify the packet flow, (ii) IP address lookup to obtain the packet forwarding information, (iii) packet filtering to implement security strategies in traffic forwarding, and (iv) a careful scheduling of the queues prior to packet transmission, to satisfy the QoS requirements. The use of active queue management schemes, that make selective and preventive packet drops to avoid link congestion, are also under debate. Classification of packets in the line cards requires completing multidimensional searches at a very high speed. Hierarchical packet classification has been proposed in the literature to reduce the search time. Lim et al. [1] in their paper ‘‘A New Hierarchical Packet Classification Algorithm’’ propose an efficient packet classification algorithm using the hierarchical approach where a hierarchical binary search tree which is free of empty internal nodes is constructed. This makes the algorithm efficient due to the absence of back-tracking and empty internal nodes. They also propose two refinements of their algorithm for reducing and avoiding the rule copy. Their proposal improves search performance without increasing the memory requirement. Ternary content-addressable memories (TCAMs) are also typically used for high-speed packet classification in routers by comparing packet header against rules in the database. To solve the issue of TCAM’s unsuitability for rules containing range fields, Bremler-Barr et al. [2] in their paper ‘‘Layered Interval Codes for TCAM-based Classification’’ propose a technique for efficient representations of range rules by taking into account that disjoint ranges can be encoded much more efficiently than overlapping ranges. The proposed scheme has been shown to significantly reduce the number of redundant TCAM entries caused by range rules. As the Internet moves to IPv6 due to exhaustion of IPv4 addresses, a large number of IP lookup algorithms have been proposed in the literature. Smiljanic and Cica [3] in their paper ‘‘Scalable Lookup Algorithms for IPv6’’ analyze the scalability of the IP lookup algorithms and their applicability to IPv6. The memory requirements of the algorithms in terms of lookup table sizes have been compared. Recommendation on using the optimal algorithm has been included as part of the study. Although packet filtering is a well-established scheme for detecting intrusions and maintaining security in the network, computation and space requirement dramatically grows with increasing complexity of filtering rules and security strategies. A critical task in packet filtering is managing the large filtering tables based on prefix and rangebased search. Neji and Bouhoula [4] in their paper ‘‘A prefix-based approach for managing hybrid specifications in complex packet filtering’’ proposed a solution based on signed prefixes to incorporate ranges in the data structures used for filtering. To show the advantages of their scheme, the authors have analyzed the worst case time and space complexity as a function of the size of the filtering table and packet header size.