LCR Solution for Performance and Cost-efficient Inter-domain Traffic Distribution Mirosław Kantor, Piotr Chołda, and Andrzej Jajszczyk Department of Telecommunications, AGH University of Science and Technology 30-059 Krakow, Poland, Telephone: (+48 12) 617–28–52, Fax: (+48 12) 634–23–72 E-mails:{kantor, cholda, jajszczyk}@kt.agh.edu.pl Abstract— In the highly connected Internet most of domains have more than one interconnection link to different providers. In such a multi-homed scenario destinations for inter-domain traffic may be reachable through multiple egress routers with different offered cost and path performance. In this paper, we address the problem of route selection in a multi-homed stub network to optimize transit costs and paths performance offered by operator to its customers. As the indicators of quality of service (QoS) related to interconnections and routes offered by interconnected partners we consider a length of AS-path (Autonomous System) and latency experienced on that path as measured by round-trip- time (RTT ). We present some QoS optimization models together with heuristic algorithms to solve the problem. To evaluate the performance of the proposed optimization models a Least Cost Routing (LCR) solution supporting optimal routing decisions has been used. The results indicate that taking into account both cost and performance is crucial for assuring the satisfied level of service for end-users with optimized cost for the operator. I. I NTRODUCTION Development of Next Generation Networks leads to a multiservice transport layer within a multi-domain environ- ment. To be able to interconnect with the entire world every Internet Service Provider (ISP) needs to have some business agreements with transit providers. As the number of potential interconnection partners increases, multihomed operators face more and more routing alternatives for sending and receiv- ing traffic. Operators also require the ability to ensure that connections in their networks are routed according to the optimal route to maximize their income and assure the required level of service performance for their customers. Although the primary purpose of multi-homing is to enhance reliability of the network, it is also desirable to use multi-homing for load-balancing and latency reduction. In this paper, we address the problem of ‘intelligent’ route selection in a multi-homed stub network in order to to optimize network latency to various destinations as measured by round-trip-time (RTT ) together with transit cost minimization. As different providers show different latencies for the same destinations, a dynamic provider selection based on delay measurements can result in significant performance improvements. However, the inter- domain traffic engineering is commonly applied today in a trial-and-error only fashion. Therefore, the need for developing algorithms supporting the choice of optimal interconnection routes becomes crucial. In the paper, we present the results of using the Least Cost Routing (LCR) solution [1]. The LCR tool, with implemented heuristic algorithms, finds the optimal route configurations for all transit traffic, assigning the most appropriate Point of Interconnections (POIs), i.e., egress routers, for each of the traffic flows. Thus, the operator obtains a network configuration table that can be manually or automatically implemented into the network control systems. The reminder of this paper is organized as follows. Sec- tion II of this paper presents an overview of papers dealing with optimization of cost and performance for multihoming. In Section III, connection model and used QoS parameters are described. Mathematical models and optimization heuristics are given is Section IV. The scenario assumptions and perfor- mance evaluation of proposed QoS optimization heuristics are presented in Section V. Section VI summarizes the work. II. RELATED WORK There are some books that cover modeling and optimization methods for problem of optimal distribution of inter-domain traffic among available interconnection partners, e.g., [2] which is an authoritative reference on communication and computer network design through mathematical optimization- oriented modeling. The LCR problem was also analyzed in some papers. The interdomain routing issues from the mechanism design point of view are presented in [3]. In [4] the problem of provisioning end-to-end bandwidth guarantees across multiple ASes is analyzed. There are also a few research studies on smart routing, e.g., [5], [6] where the focus of these studies is on network performance. User costs being another major incentive to use multihoming, are considered, e.g., in [7] where a series of novel smart routing algorithms to optimize the cost and performance for multihomed users was proposed. III. CONNECTION MODEL A general considerd connection model is presented in Fig- ure 1. A multi-domain communication network is represented by a set of POI nodes (egress routers) which represent the gateways between domains and a set of abstract nodes p ∈ P corresponding to interconnection partners. Some parameters are related to each POI, e.g., the tariff (related to POI inter- faces) and the capacity. By the capacity in this context we mean the maximum traffic volume which can be sent through the partner, POI or interface during a considered time period (e.g., a month, a day, an hour, etc.). This limitation can result from technical constraints of particular locations (e.g., the interconnection link bandwidth). 978-1-4244-6705-1/10/$26.00 ©2010 IEEE