Alexander Gersht

Joint optimization of data network design and facility selection

The authors describe a data network design model based on a mixed integer/linear programming (MILP) formulation that does not, as do most other approaches, separate link capacity and facility selection from routing and topological design; it fully integrates these processes to capture the important couplings that exist between them. The performance constraints are incorporated into the model in such a way that they are linear, but lead to the same grade of service for a balanced network as nonlinear average network delay constraints. It is shown that this formulation leads to a natural decomposition of the optimal design problem into two subproblems solvable sequentially. In the absence of capacity allocation constraints, the capacity and flow assignment problem is solved optimally and efficiently as part of the overall design process. Moreover, the model leads directly to the solution of multifacility design problems. A fast link reduction algorithm that efficiently designs single or multifacility networks and yields robust local extrema is presented. This algorithm is based on a special-purpose monotonic greedy drop heuristic procedure. An important application of this model is the design of multifacility networks. >

A congestion control framework for ATM networks

A congestion control framework for ATM (asynchronous transfer mode) networks is proposed. Specifically, it is suggested that the network provide two different services: express service and first-class service. Express service is appropriate for real-time applications (e.g., voice and video) whereas first-class service is appropriate for non-real-time applications (e.g., data). To provide such services, the proposed congestion control scheme prevents congestion inside the network by controlling the congestion at two levels: virtual circuit (VC) level and packet level. Express VC traffic is not subject to any flow control. At each intermediate node, transit packets are simply relayed with no traffic control. Various issues related to this scheme are discussed and its performance analyzed.<<ETX>>

Optimal routing in circuit switched communication networks

We consider the optimal circuit routing problem. The problem consists of accommodating a given circuit demand in an existing circuit-switched network. The objective is to find a circuit accommodation providing the maximum residual capacity over the network under the total circuit cost constraints. Practical considerations require a solution which is robust to the variations in circuit demand and cost. The objective function for the circuit routing problem is not a smooth one. In order to overcome the difficulties of nonsmooth optimization, a sequence of smooth convex optimization problems is considered. The optimal algorithm for the circuit routing problem is obtained as a limiting case of the sequence of the optimal routing strategies for the corresponding smooth optimization problems. The proof of its convergence to the optimal solution is given. This optimization algorithm is capable of efficiently handling networks with a large number of commodities. It also satisfies the above-mentioned robustness requirements. Numerical results are discussed.

Dynamic bandwidth-allocation and path-restoration in SONET self-healing networks

This paper presents a new scheme for real-time bandwidth allocation and path restoration (BARS) in mesh networks via SONET wideband digital cross-connect systems (WDCSs) in response to demand and load dynamics and link and/or node failure(s). The scheme dynamically maximizes bandwidth allocation while ensuring full service restorability. Since the physical network capacity is limited, sometime not all the demand can be accommodated under the full restorability requirement. This demand in SONET BARS is rejected fairly at the network boundary even if capacity for allocation is available. Bandwidth allocation and fair demand admission are optimized jointly under the full restorability requirements. The implementation of SONET WDCS does not need excessive storage. An efficient parallel algorithm for solving the optimization problem is also presented. The algorithm produces superior spare capacity assignments compared to the results in the literature.

Optimal dynamic virtual path bandwidth allocation and restoration in ATM networks

This paper presents an optimal dynamic virtual path (VP) bandwidth allocation and restoration scheme for mesh ATM networks (in the context of the layered traffic control architecture). The scheme integrates the dynamics of demand admission, VP bandwidth allocation, and logical spare capacity assignment for maximizing network throughput while ensuring full traffic restorability. We present an optimal parallel algorithm that minimizes the total rejected bandwidth demand and cell loss while satisfying the maximal cell loss, delay, and 100% restorability requirements for a given set of failure scenarios. The optimal spare capacity assignment in the presented approach follows directly from the network admission and bandwidth allocation decisions. The algorithm also equalizes cell losses on the VPs thus providing cell-level fairness.

Dynamic Bandwidth Allocation, Routing, and Access Control in ATM Networks

We propose a layered architecture for adaptive traffic control in ATM networks that utilizes the virtual path (VP) concept. The proposed control guarantees quality of service and simplifies call and cell processing. The control is organized in three levels: a VP level, a virtual circuit (VC) level, and a cell level.

Performance management in SONET-based multi-service networks

A performance management scheme is introduced to satisfy end-to-end performance requirements of multiclass services, using a nested control scheme. The network-level controller periodically computes an optimal bandwidth allocation policy for the entire network, and downloads parameters to the call- and packet-level controllers to regulate network traffic accordingly. The call-level controller limits SONET (synchronous optical network) pipe fills, exercises access-control fairly across source destination pairs, and is based on profitability across call classes. The packet-level controller provides congestion-free high-speed data transmission within the SONET pipes designated and realized by the network and call controllers, respectively.<<ETX>>

A new algorithm for the solution of the minimum cost multicommodity flow problem

The multicommodity minimum cost flow problem (MMCF) is to determine a minimum cost multicommodity flow through the constrained network. This paper considers a new barrier-penalty optimization algorithm for the solution of a general linear MMCF problem. The algorithm does not require an initial point to be feasible. This algorithm is computationally efficient and allows one to solve MMCF problems with a large number of commodities. It is also applicable to the convex MMCF problems with nonlinear constraints and/or objective function. The proof of its convergence and a new algorithm for calculating a lower bound are presented. The computer implementation of the algorithm is discussed, and computational experience is reported.

Real-time traffic management by a parallel algorithm

The joint access-control and routing problem for distributed networks is formulated as an equilibrium programming problem (EPP). The strategy, for an upcoming control period, decreases call blocking and balances network load by maximizing network residual capacity, and fairly rejects at the source the source-destination demands that are expected to exceed network capacity. The EPP formulation allows for both decentralized implementation of the joint access control and routing problem and massive parallelization of the optimization procedure to satisfy real-time requirements. The computational complexity of the algorithm decreases proportionally with the number of processors used. The convergence of the decentralized algorithm to the network-wide optimum is proved. Compared to the case where subnet EPPs are solved independently by each controller, simulation results show significantly better utilization of network resources when subnet controllers jointly solve the network-wide EPP. Performance analysis of the algorithm and numerical studies show that the algorithm is well suited for real-time implementation in large networks or internetworks. >

Architecture for restorable call allocation and fast VP restoration in mesh ATM networks

This paper presents an architecture for restorable call allocation and fast virtual path (VP) restoration in mesh ATM networks. In this architecture, virtual working and spare capacities needed for call allocation and restoration are reserved and released dynamically on a call-by-call basis at the time of call admission and termination. The architecture does not require advance assignment of spare and working capacities. To shorten the call processing delay, this is done in a parallel-distributed fashion. To provide restorable call allocation, parallel-distributed call processing algorithms of the sender-chooser type are suggested. The algorithms integrate on the call level virtual bandwidth allocation, virtual spare capacity assignment with fixed, alternate, or state-dependent routing. Each routing scheme leads to a particular tradeoff between call processing complexity, call setup delay, and bandwidth efficiency. For each pair of nodes, two sets of VPs are provisioned. The first is a working VP (WVP) set, to be used for call allocation during the normal operation. The second is a spare VP (SVP) set, to be used for VP restoration in the event of failure of network elements (NE). Each SVP protects a preassigned subset of the node pair WVPs. Each SVP is selected to be link/node disjoint from the WVPs that it is assigned to protect. This assures protection of the WVP set by a small number of SVPs. Since SVPs are preset and appropriate virtual spare capacities are reserved in advance, the architecture guarantees full restorability and provides very fast restoration. The restoration is done on the VP level in a self-healing manner. The suggested architecture requires only local information to be maintained at each node.