Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 5

Genetic Algorithm for Survivable Network Design Problem with Simultaneous Unicast and Anycast Flows

Abstract . Given a connected, weighted, undirected graph G = (V, E), a set of nodes, a set of links with modular cost based on ACMC model [2] and a set of customers demands. This paper proposes Genetic Algorithm for solving the ACMC Survivable Network Design Problem (A-SNDP). The goal is to design connections based on customers demands with the smallest network cost to protect the network against all failures. This problem is NP-hard. The experimental results are reported to show the efficiency of proposed algorithm comparing to the Tabu Search algorithm [2]. Keywords: Survivable Network Design, Anycast, Unicast, Genetic Algorithm.

Introduction
In the recent years, there are many kinds of transmissions have been used, included unicast and anycast flow, which are two of the most popular types. A transmission connected one host to another is called unicast. An anycast is defined as one-to-oneof-many transmission to deliver a packet to one of many host, which is applied in Domain Name Service (DNS), Web Service, Overlay Network, peer-to-peer (P2P) systems, Content Delivery Network (CDN), software distribution[2]. The popularity of anycast technology will increase in the near future, since many new services that can use anycast paradigm are developed [1]. The impact of Internet and other computer networks can be seen in almost all areas of our lives, including business, science, technology, so on, and their importance will grow. Thus, any network failure can cause serious consequences affect to many people and corporations. In [12], a single link failure may affect more than 30,000 users and it takes 12 hours to fix. Therefore, designing the survivable network is a crucial problem. In network with simultaneous unicast and anycast flow, the objective is to minimize the network cost required to protect the network against failure. Because we solve the SNDP problem based on ACMC (All Capacities Modular Cost) model which is defined in [2], we call this problem by A-SNDP (ACMC Survivable Network Design Problem). A-SNDP problem is defined as the following: Given an undirected graph G = (V, E). Each link divides into some bandwidth levels, each level has a corresponding cost. Link cost is the cost of total of corresponding bandwidths using in this link. And network cost is total of link cost. A connection from a node to another is a unicast transmission. An anycast is also a connection from a node to another, but the difference is that the destination node has a replica server which backs up for it. In this problem, a demand required by customer is a connection between two nodes with corresponding bandwidth and type of connection. The goal of problem is to find a set of connections for all demands such that the network cost is minimum.

We can formulate the problem as following: Find an appropriate connection for each customers demand so as to minimize:

NCost = ci
i

(1)
k

With condition is that if ( Bk 1 <

Rij < B
j

) then ci = Ck, where ci is cost

of link i, Bk, Ck is bandwidth and corresponding cost in index k. To solve A-SNDP, we have to build a set of path for all demands to minimize the network cost. To guarantee the survivability of connections, we also use backup path approach [2] [3] [4] [5]. In particular, each connection (anycast or unicast) is divided into working path and backup path (they can use some same links, but not completely). If the working path is broken, the backup path must be restored and conversely. In order to minimize the network cost, we propose an approach based on Genetic Algorithm with a new way of encoding. After 50 running times of 10 test sets, our proposed algorithm has better result comparing to Tabu Search [2]. The rest of this paper is organized as following: Section II describes the related works. Our new approach to encode solution is showed in section III. In section IV, we present the proposed algorithms to solve A-SNDP. Our experiments and computational and comparative results are given in section V. The paper concludes with section VI with some discussions on the future extension of this work.

Related Works
The SNDP is generally presented in [10], when considering both economics and reliability in telecommunication network design. Thus, there are two problems requested in the A-SNDP. Those are to guarantee the survivability of network system and to minimize the network cost. The most popular way mentioned in many researches is the single backup path approach. The main idea of this method is as following: Each connection has a working path and a backup path. The working path is used for transmitting data in normal, failure-free state of the network. After a failure of the working path, the failed con-nection is switched to the backup path [2], [3], [4], [5]. In the literature there are many papers on minimizing the network cost for SNDP problem (see [2], [8], [10] and references therein). They use branch and bounds or branch and cut methods to find optimal precisely solutions. These methods can only use for small networks with 30 nodes and 100 edges. For larger network, they may propose evolutionary algorithms, tabu search [2], and simulated annealing [8]. In [10], Vissen and Gold apply the evolution strategy (ES). Following the discussion of [10], the quality of result is very flexible and effective. Overall, the graph instance with 30 nodes and 200 edges (the (30, 200) ES) with discrete recombination delivers the best solution. The (10, 50) ES that operated with a smaller population size uses much less resources while still delivering good cost results, but it is not as good

as the (30, 200) ES. Clearly, when using ES, a larger population helps to achieve a better result than a smaller one by avoiding or delaying convergence on local suboptimal. However, this algorithm is useful in the network which has only unicast flows. With the network which has both anycast and unicast flows, K.Walkowiak and Jacub Gladysz [14] presented a heuristic algorithm for solving. The main idea of this algorithm is based on Flow Deviation method [8] and Local Search algorithm [15] They achieve the quite good result with Polska (12 nodes, 36 links, 65 unicast, 12 anycast) network, detail that the average gap of the propose heuristic to optimal results is 7.11%. Furthermore, K. Walkowiak and Jakub Gadysz [2] have built Tabu search algorithm based on hill climbing algorithm with some heuristics to solve this problem. They experimented this algorithm with three large instances which are Polska (12 nodes, 36 links, 65 unicast, 12 anycast), Germany (17 nodes, 52 links, 119 unicast, 13 anycast), Atlanta (26 nodes, 82 links, 234 unicast, 22 anycast) and received effective results. In particular with Polska network, they achieve the average gap to optimal results is 2.57% for 70% anycast/30% unicast case and 2.00% for 80% anycast/ 20% unicast case. However, their Tabu Search algorithm is quite simple and their results cannot be optimal completely. In the next section, we introduce two algorithms for solving A-SNDP. We hope that our proposed algorithms will have a better result than previous algorithms.

Path Database based Encoding(PDE)


1.1 1.2 Definition PDE for Solving ASNDP

Genetic Algorithm using PDE


1.3 1.4 1.5 1.6 1.7 Individual Representation Selection Crossover operator Mutation operator Proposed algorithm

Experimental Results
1.8 Problem Instances

In our experiment, we used three real world instances. They are Polska, Germany and Atlanta network [2][14]. With each instance, we randomly create 10 test sets which are different from the content of customers demands. Table 1 shows three real world instances.
Table 1. REAL WORLD PROBLEM INSTANCES

Instances Polska Germany Atlanta

Nodes 12 17 26

Edges 36 52 82

Unicast 40 44 74

Anycast 12 13 22

We install our proposed algorithms and compare its performance with Tabu Search in [2]

1.9

System Setting

In the experiment, the system was run 50 times for each test set. All the program were run on a machine with Intel Core 2 Duo U7700, RAM 2GB, Windows 7 Ultimate, and were installed by C++ and Java language. 1.10 Computational Results

The experiment shows that:

Conclusion Reference
Krzysztof Walkowiak, A Flow Deviation Algorithm for Joint Optimization of Unicast and Anycast Flows in Connection-Oriented Networks. In: Computational Science and Its Applications ICCSA 2008, pp. 797807, Springer, Perugia, Italy (2008). Jakub Gadysz, Krzysztof Walkowiak, Tabu Search Algorithm for Survivable Network Design Problem with Simultaneous Unicast and Anycast Flows, Intl Journal Of Electronics And Telecommunications, VOL. 56, NO. 1, pp. 4148 (2010). Krzysztof Walkowiak , A New Function for Optimization of Working Paths in Survivable MPLS Networks. In: Computer and Information Sciences ISCIS 2006, pp. 424--433, Springer, Istanbul (2006) W. Grover, Mesh-based Survivable Networks: Options and Strategies for Optical, MPLS, SONET and ATM Networking, Prentice Hall PTR, Upper Saddle River, New Jersey (2004). V. Sharma, F. Hellstrand, Framework for MPLS-based recovery, RFC 3469 (2003). J. Vasseur, M. Pickavet, P. Demeester, Network Recovery: Protection and Restoration of Optical, SONET-SDH, IP and MPLS, Morgan Kaufmann, San Francisco (2004). A. Kasprzak, Algorithms of flow, capacity and topology structure in computer networks, Monography, Wroclaw, Polish (1989). M. Pioro, D. Medhi, Routing, Flow, and Capacity Design in Communication and Computer Networks, Morgan Kaufmann Publishers (2004). K. Walkowiak, Anycast Communication A New Approach to Survivability of ConnectionOriented Networks. In: Communications in Computer and Information Science, pp. 378389, Springer, Berlin (2003). Volker Nissen, Stefan Gold, Survivable network design with an evolution strategy. In: Success in Evolutionary Computation, pp. 263283, Springer, Berlin (2008). Johnson & Deering, Reserved IPv6 Subnet Anycast Addresses, RFC 2526, March 1999. Mukherjee, B., Optical WDM networks, Springer, Berlin (2006). Anycast vs Unicast, http://communitydns.eu/Anycast.pdf Jakub Gladysz, Krzysztof Walkowiak: Optimization of survivable networks with simultaneous unicast and anycast flows. ICUMT 2009, pp. 1-6

You might also like