New Dynamic Network Design and Provisioning Algorithms for Broadband Connection Services Considering Fairness Masahiro Nakagawa 1 , Hiroshi Hasegawa 1 , Ken-ichi Sato 1 , Ryuta Sugiyama 2 , Tomonori Takeda 2 , Eiji Oki 2* , and Kohei Shiomoto 2 1 Nagoya University, Furo-cho, Chikusa-ku, Nagoya, 464-8603 Japan m_nakaga@echo.nuee.nagoya-u.ac.jp, {hasegawa, sato}@nuee.nagoya-u.ac.jp 2 NTT Network Service Systems Laboratories, 3-9-11 Midori, Musashino, Tokyo, 180-8585 Japan {sugiyama.ryuta, takeda.tomonori, oki.eiji, shiomoto.kohei}@lab.ntt.co.jp Abstract— We propose novel dynamic network control algorithms that reduce the blocking probability for dynamic bandwidth service provisioning. The expected network services include on-demand broad bandwidth provisioning services and layer one VPN. The important service attribute of fairness in terms of path length is effectively achieved by introducing a simple weighting function that considers path length and link utilizations of intermediate links of candidate paths. The algorithm achieves enhanced network utilization by rerouting existing paths to alternative routes without disruption. Numerical examples demonstrate that the developed algorithms attain not only a high degree of fairness but also low service blocking probability. I. INTRODUCTION Due to the rapid penetration of broadband access, Internet traffic has been exploding throughout the world. In order to cope with this large increase in traffic demand, new optical transport systems employing wavelength routing via ROADMs (Reconfigurable Optical Add/Drop Multiplexers) have been widely deployed [1], [2]. Demands for IP/Ethernet-based (layer 2 and 3) virtual private network services are also rapidly increasing. This fuels the advancement of network control technologies based on ASON/GMPLS [3], [4]. Such developments are spurring carriers into providing new layer one services that offer dynamic and adaptive bandwidth for the creation of the cooperative utility backbone and that are necessary for wholesale carrier business [5], [6]. They include broad bandwidth on-demand provisioning services such as Optical Mesh Service provided by AT&T [7] and JiT (Just in Time) service by Verizon [5]. These VPN services require agile network reconfigurability. Ultra-high definition video (raw bit rate of 72 Gbps) and 4-k cinema (6 Gbps) distribution, and Grid-computing will also become the key services. Dynamic and adaptive bandwidth provisioning capability is a key to satisfy such broadband requests economically. The connection requests are likely to be more schedule-based (unlike sporadic telephone calls), and guaranteed bandwidths (in other words, guaranteed quality) need to be provided [5], [6]. To enable such emerging large-capacity services cost effectively and to meet diverse service requirements, novel dynamic network control technologies need to be developed. In the current approach to the provisioning of dynamic services, minimization of service blocking probability or good load balancing is taken as the criterion of network control. Based on this criterion, a path (in this paper, path is used interchangeably with circuit, since the service will, for example, use VC-3/4; SONET/SDH higher-order paths) is setup when a connection request is received. A comprehensive review of lightpath establishment can be found in [8]. Generally speaking, the routes of existing paths will not be changed while a connection is being made. However, since the traffic distribution keeps changing all the time, path assignment will slowly diverge from the optimal path accommodation. What is worse, it may not be possible to route new paths through congested areas. Re-optimization by rerouting existing paths in response to new demands is, therefore, necessary to attain lower blocking probabilities (higher network utilization). The impact of rerouting has been discussed for circuit-switched telephone networks [9], [10]. It has also been recently introduced to optical WDM networks [11]-[14]. In [15], an analysis of rerouting in circuit-switched networks was provided; it verified that rerouting can significantly increase throughput compared to traditional dynamic routing. Rerouting for the provisioning of multi-granularity connections in optical WDM networks was also studied [14]. Some rerouting techniques were summarized in [16]. Rerouting is a simple operation that switches an existing path from its current route to another route. We can classify it into two strategies; passive rerouting and intentional rerouting [13]. Passive rerouting means reroute existing paths to accommodate new path requests which would otherwise be blocked. The basic idea of intentional rerouting is to intentionally reroute some existing paths to vacant routes in advance if it yields better load balancing. Latest studies focused not only on reducing the blocking probability or on achieving better load balancing, but also on reducing disruption time. For example, a rerouting scheme called move-to-vacant wavelength-retuning (MTV-WR) [11] has been proposed and achieves shorter disruption time and further studies have been published [12]-[14]. Most of the rerouting schemes developed so far generally cause path disruption; an existing path is disconnected before the alternative path is setup. Such disruption can significantly impair the quality of real-time applications such as live video *Eiji Oki is now with University of Electro-Communications, 1-5-1 Chofugaoka, Chofu, Tokyo, 182-8585 Japan (e-mail: oki@ice.uec.ac.jp).