Chapter 1 Introduction
n recent years, synchronous optical network (SONET) ring networks have been widely deployed for the optical network infrastructure. The progresses of optical networks evolve with time; meanwhile, the carried traffic streams surge. The transmission capacity of an optical network largely increases because of the development and application of wavelength division multiplexing (WDM) technologies which allow multiple wavelengths to be delivered in a single fiber at the same time [1]. Besides, the capacity of one wavelength has grown up to optical-carrier (OC) 192, approximately 10 Gb/s.
A traffic stream is normally assigned in one wavelength. However, the required bandwidth of a traffic stream, called lower-speed traffic stream, is much smaller than the bandwidth capacity of a wavelength. If every lower-speed traffic stream occupies an individual wavelength, bandwidth waste is an obvious and serious drawback.
In order to efficiently utilize the network resources, many lower-speed traffic streams can be multiplexed onto a high-speed wavelength by traffic-grooming technique. This technique is performed by Add/Drop Multiplexer (ADM) equipment. An overview of the traffic-grooming technique and survey of some typical works were reported in [2].
In the design of traffic-grooming, three issues to be considered: network configuration,
I
the mesh network, which their backbone is a long-haul backbone network. In the early stage of optical network, the ring network is employed in a long-haul backbone network. With the increasing of the traffic load, the mesh network has been studied by many researches because the mesh network sustains more traffic load than the ring network. However, for metro networks, ring network is preferable because the mesh network is an over-provisioning and expensive solution. Instead, many people study the ring network for the metro networks.
For the traffic characteristics, two properties are used to describe the behavior of traffics in this thesis. One is bandwidth request and the other one is connection pattern. Traffics may be categorized into uniform or non-uniform ones according to their required bandwidth. The bandwidth request is the same for all the users, called uniform traffic; otherwise, it is non-uniform traffic. And, the traffics may also be categorized into static or dynamic ones according to system connection including the amount of connection, the occupied bandwidth of all connection. The system connection does not change with time for static traffic while it does for dynamic traffic. Notably, the required bandwidth of each call request does not change during its transmission time in this thesis.
Besides, many objectives were proposed for the definitions of cost functions and may be categorized into five types. The first one is minimization of the required number of Line Terminating Equipment (LTE) over all the network nodes. The second one is reduction of the cost of actual electronic processing involved in Optical-Electric-Optical (OEO) routing or minimization of the average hop counts. The third one is minimization the maximum number of lightpaths originating/terminating at a network node. The fourth one is maximization of traffic throughput of the optical network. The last one is minimization of the blocking probability of the new call request in dynamic traffic.
In [3], the traffic-grooming problem was considered WDM mesh network with static and uniform traffic. The new node architecture was proposed for a WDM mesh network with traffic-grooming capacity and cost function was designed to maximize the total
successfully-routed lower speed traffic. A traffic-grooming problem in WDM ring was studied in [4], where the traffic is static and non-uniform. The authors presented a framework of bounds, both upper and lower bounds, on the optimal value of the amount of traffic electrically routed. In [5], a traffic grooming problem on general topology network was studied, where the objective of cost function was minimization of the number of required wavelengths. In this work, an integer linear programming problem was formulated for traffic grooming, and a heuristic solution was proposed. In [6], the authors considered a Min-Max objective, which was desirable to minimize the cost at the node where this cost is maximum.
They presented a polynomial-time traffic grooming algorithm for minimizing the maximum electronic port cost in both uni- and bi-directional ring.
The ring network was considered with static and non-uniform traffic [7]. The simulated-annealing-based heuristic algorithm for the traffic grooming problem was proposed for minimization of the total cost of electronic equipment. Both nui- and bi-directional ring were considered in [8]. It considered two kinds of traffic. One was static and uniform traffic and the other was static and non-uniform traffic. For minimizing number of wavelengths, the greedy heuristic algorithm was employed for two kinds of traffic. The author also proposed another heuristic algorithm for static and non-uniform traffic according to the same objective. In [9], the ring network was considered with static and uniform traffic.
The author proposed the heuristic algorithm for minimizing number of wavelengths and also considered computational complexity.
The network architecture of [10] was ring network and two kinds of traffic are considered:
static and uniform traffic, static and non-uniform traffic. The heuristic algorithm was
ring was employed under the static and uniform traffic. For minimizing the number of ADMs, the tighter lower bounder is derived. The author also proposed the heuristic algorithm.
Although the ring network has been studied by many researchers, dynamic and non-uniform traffic has not considered yet. Therefore, we are motivated to study the traffic grooming problem in metro-access ring network with dynamic and non-uniform traffic. In the thesis, the metro-access ring network has nodes in WDM-PON architecture connecting to end-users. We investigate the node architecture from the current equipments and propose the node architecture which is equipped in metro-access ring network and have the ability of traffic grooming. The objective of this work is to effectively reduce the new call blocking rate and maximize the utilization efficiency of the wavelength used.
Integer linear problem (ILP) methodology is applied to formulate for the traffic grooming problem in this thesis. And, in order to solve the ILP, we first propose a simulated annealing-based traffic grooming algorithm (STGA) to obtain the optimal solution. However, the STGA algorithm is infeasible due to its computation complexity. Alternatively, we propose a heuristic algorithm, called a heuristic-based traffic grooming algorithm (HTGA).
Although the HTGA algorithm is a suboptimal solution, the computation complexity of the HTGA is much less than that of the STGA.
For our objective, we consider grooming traffic into the present lightpaths. If the new call request can not be arranged by grooming traffic, employing the unused resource is considered. Besides, in order to more effectively increase the utilization efficiency of the wavelength used, the current lightpaths are sorted in descendent order according to the value of utilization of lightpaths. Based on above, there are the three main operations in HTGA algorithm: operation of traffic grooming, operation of wavelength assignments, operation of traffic rearrangement. The operations of traffic grooming and wavelength assignments may be executed when the new call request arrives; while the operation of traffic rearrangement is executed when the required large bandwidth of a call leaves. The main purpose of these operations is to achieve the better system utilization without changing the lightpath topology
as could as possible. And, the blocking rate of the new call request is also reduced by HTGA algorithm.
Simulation comparisons between the two algorithms were conducted in the thesis for performance evaluation. The results were demonstrated in terms of system utilization and computation complexity. In the system utilization, the HTGA algorithm is only 10% less than the STGA algorithm. And, the computation complexity of the STGA algorithm is much larger than that of the HTGA algorithm. In addition, the number of rearranged lightpaths by the HTGA algorithm is fewer than the ones by the STGA algorithm. From these results, we also point out that the STGA algorithm is infeasible with the current equipments.
The thesis is organized as follows. In Chapter 2, we describe the network architecture, the node architecture. The detail of the traffic assumption is introduced in the source model of Chapter 2. In addition, the problem is mathematically formulated for the optical ring network with dynamic traffic model. In Chapter 3, the optimal STGA algorithm and suboptimal HTGA algorithm are proposed and the operations of two algorithms are described. The procedures of two algorithms are also revealed in detail in Chapter 3. The simulation results of two algorithms are discussed in chapter 4. At last, Chapter 5 concludes the thesis.