Hisao Kameda

Optimal dynamic mobility management for PCS networks

We study a dynamic mobility management scheme: the movement-based location update scheme. An analytical model is applied to formulate the costs of location update and paging in the movement-based location update scheme. The problem of minimizing the total cost is formulated as an optimization problem that finds the optimal threshold in the movement-based location update scheme. We prove that the total cost function is a convex function of the threshold. Based on the structure of the optimal solution, an efficient algorithm is proposed to find the optimal threshold directly. Furthermore, the proposed algorithm is applied to study the effects of changing important parameters of mobility and calling patterns numerically.

Optimal load balancing in distributed computer systems

An important consideration in improving the performance of a distributed computer system is the balancing of the load between the host computers. Load balancing may be either static or dynamic; static balancing strategies are generally based on information about the systems average behavior rather than its actual current state, while dynamic strategies react to the current state when making transfer decisions. Although it is often conjectured that dynamic load balancing outperforms static, careful investigation shows that this view is not always valid. Recent research on the problem of optimal static load balancing is clearly and intuitively presented, with coverage of distributed computer system models, problem formulation in load balancing, and effective algorithms for implementing optimization. Providing a thorough understanding of both static and dynamic strategies, this book will be of interest to all researchers and practitioners working to optimize performance in distributed computer systems.

An algorithm for optimal static load balancing in distributed computer systems

The authors propose a load-balancing algorithm that determines the optimal load for each host so as to minimize the overall mean job response time in a distributed computer system that consists of heterogeneous hosts. The algorithm is a simplified and easily understandable version of the single-point algorithm originally presented by A.N. Tantawi and D. Towsley (1985). >

Load balancing problems for multiclass jobs in distributed/parallel computer systems

Load balancing problems for multiclass jobs in distributed/parallel computer systems with general network configurations are considered. We construct a general model of such a distributed/parallel computer system. The system consists of heterogeneous host computers/processors (nodes) which are interconnected by a generally configured communication/interconnection network wherein there are several classes of jobs, each of which has its distinct delay function at each host and each communication link. This model is used to formulate the multiclass job load balancing problem as a nonlinear optimization problem in which the goal is to minimize the mean response time of a job. A number of simple and intuitive theoretical results on the solution of the optimization problem are derived. On the basis of these results, we propose an effective load balancing algorithm for balancing the load over an entire distributed/parallel system. The proposed algorithm has two attractive features. One is that the algorithm can be implemented in a decentralized fashion. Another feature is simple and straightforward structure. Models of nodes, communication networks, and a numerical example are illustrated. The proposed algorithm is compared with a well-known standard steepest-descent algorithm, the FD algorithm. By using numerical experiments, we show that the proposed algorithm has much faster convergence in terms of computational time than the FD algorithm.

Braess-like paradoxes in distributed computer systems

We consider optimal distributed decisions in distributed computer systems. We identify a Braess like paradox in which adding capacity to the system may degrade the performance of all users. Unlike the original Braess paradox, we show that this behavior occurs only in the case of finitely many users and not in the case of infinite number of users.

Mixed equilibrium (ME) for multiclass routing games

We consider a network shared by noncooperative two types of users, group users and individual users. Each user of the first type has a significant impact on the load of the network, whereas a user of the second type does not. Both group users as well as individual users choose their routes so as to minimize their costs. We further consider the case that the users may have side constraints. We study the concept of mixed equilibrium (mixing of Nash equilibrium and Wardrop equilibrium). We establish its existence and some conditions for its uniqueness. Then, we apply the mixed equilibrium to a parallel links network and to a case of load balancing.

Equilibria for multiclass routing in multi-agent networks

We study optimal static routing problems in open multiclass networks with state-independent arrival and service rates. Our goal is to study the uniqueness of optimal routing under different scenarios. We consider first the overall optimal policy that is the routing policy whereby the overall mean cost of a job is minimized. We then consider an individually optimal policy whereby jobs are routed so that each job may feel that its own expected cost is minimized if it knows the mean cost for each path. This is related to the Wardrop equilibrium concept in a multiclass framework. We finally study the case of class optimization, in which each of several class of jobs tries to minimize the averaged cost per job within that class; this is related to the Nash equilibrium concept. For all three settings, we show that the routing decisions at optimum need not be unique, but that the utilizations in some large class of links are uniquely determined.

Paradoxes in distributed decisions on optimal load balancing for networks of homogeneous computers

In completely symmetric systems that have homogeneous nodes (hosts, computers, or processors) with identical arrival processes, an optimal static load balancing scheme does not involve the forwarding of jobs among nodes. Using an appropriate analytic model of a distributed computer system, we examine the following three decision schemes for load balancing: completely distributed, intermediately distributed, and completely centralized. We show that there is no forwarding of jobs in the completely centralized and completely distributed schemes, but that in an intermediately distributed decision scheme, mutual forwarding of jobs among nodes is possible, leading to degradation in system performance for every decision maker. This result appears paradoxical, because by adding communication capacity to the system for the sharing of jobs between nodes, the overall system performance is degraded. We characterize conditions under which such paradoxical behavior occurs, and we give examples in which the degradation of performance may increase without bound. We show that the degradation reduces and finally disappears in the limit as the intermediately distributed decision scheme tends to a completely distributed one.

A Finite-Source Queue with Different Customers

finite-source queuing model (sometimes called the finite-population, machine-interference, or machine-repairman model), which has often been used in analyzing time-sharing systems and multi- programmed computer systems, is invesugated. The model studied here has two service staUons, a processor (single server) and peripherals (infinite server), and a finite number of customers (or jobs) that have a distract service rate at the processor. The model is in eqmhbnum. It is shown that the utilization factor of the processor can be obtained in an analyuc form and ts independent of various scheduling disciphnes employed at the processor, such as FCFS, generahzed processor sharing, preempUve (resume) and nonpreemptwe priority disciphnes, under some condiaon. Other relevant propemes of this model are also shown. The range within which these properties hold is discussed, and some examples are given. Examples of appficatlon to multiprogrammmg and tune-sharing systems are given; in particular, It Is shown that the often used dynamic dispatching pohcy (which gwes the higher preempuve priority to the more I/O oriented job) is optimal within the framework of this multiprogramming model. Categories and SubJect Descriptors:

A performance comparison of dynamic vs. static load balancing policies in a mainframe-personal computer network model

Distributed computer systems can share job processing in the event of overloads. Load balancing involves the distribution of jobs throughout a networked computer system, thus increasing throughput without having to obtain additional or faster computer hardware. Load balancing policies may be either static or dynamic. Static load balancing policies are generally based on the information about the average behavior of system; transfer decisions are independent of the actual current system state. Dynamic policies, on the other hand, react to the actual current system state in making transfer decisions. This makes dynamic policies necessarily more complex than static ones, and truly optimal dynamic policies are known only for special systems. This study focuses on performance comparison between static and dynamic load balancing policies in a distributed computer system where truly optimal solutions of both dynamic and static policies have been characterized. The system consists of two types of service facilities, a mainframe node and an unlimited number of personal computer nodes. The results suggest that, in the model examined, the dynamic policy outperforms the static one in the mean response time, at most about 30 percent and for the range of parameter values such that the arrival rate is near the processing rate of the mainframe.