Thesis (Ph.D)--Industrial and Systems Engineering, Georgia Institute of Technology, 2010. / Committee Chair: Ayhan, Hayriye; Committee Member: Foley, Robert D.; Committee Member: Goldsman, David M.; Committee Member: Reed, Joshua; Committee Member: Zwart, Bert. Part of the SMARTech Electronic Thesis and Dissertation Collection.

Fluid networks are useful tools for analyzing complex manufacturing environments especially in semiconductor wafer fabrication. The makespan of a fluid network is defined as the time to drain the system, when there is fluid present in the buffers initially. Based on this definition, the question of determining the allocation of resources so as to minimize the makespan of a fluid network is known as the makespan problem. In the deterministic version of the makespan problem, it is assumed that the parameters of the system, such as incoming rates, service rates and initial inventory, are known deterministically. The deterministic version of the makespan problem for reentrant lines and multiclass fluid networks has been investigated in the literature and an analytical solution for the problem is well-known. In this work, we provide another formulation for the deterministic makespan problem and prove that the problem can be solved for each station separately. Optimal solutions for the deterministic makespan problem have been used as a guide to develop heuristics methods to solve makespan scheduling problem in the job-shop context in the literature. This provides one motivation for further investigation of the fluid makespan problem. In this work our main focus is solving the makespan problem when the problem parameters are uncertain. This uncertainty may be caused by various factors such as the unpredictability of the arrival process or randomness in machine availability due to failures. In the presence of parameter uncertainty, the decision maker's goal is to optimally allocate the capacity in order to minimize the expected value of the makespan. We assume that the decision maker has distributional information about the parameters at the time of decision making. We consider two decision making schemes. In the first scheme, the controller sets the allocations before observing the parameters. After the initial allocations are set, they cannot be changed. In the second scheme, the controller is allowed a recourse action after a data collection process. It is shown that in terms of obtaining the optimal control, both schemes differ considerably from the deterministic version of the problem. We formulate both schemes using stochastic programming techniques. The first scheme is easier to analyze since the resulting model is convex. Unfortunately, under the second decision scheme, the objective function is non-convex. We develop a branch-and-bound methodology to solve the resulting stochastic non-convex program. Finally, we identify some special cases where the stochastic problem is analytically solvable. This work uses stochastic programming techniques to formulate and solve a problem arising in queueing networks. Stochastic programming and queueing systems are two major areas of Operations Research that deal with decision making under uncertainty. To the best of our knowledge, this dissertation is one of the first works that brings these two major areas together.

published_or_final_version / Electrical and Electronic Engineering / Master / Master of Philosophy

published_or_final_version / Mechanical Engineering / Master / Master of Philosophy

In this study, I interviewed 57 low-income urban fathers about how they distribute resources between children, how they define responsible fatherhood and how they negotiate state surveillance. First, using queuing theory, I find that these fathers do not distribute their resources of time and money equally but instead give more of their resources to a smaller number of children in order to maximize their impact. I identify nine criteria that men use to prioritize among their children: timing of life course interruptions, distance, formal child support, desirability of the pregnancy, restraining orders, other resources available to the child, age of the child, gender of the child, and the child's reaching out behavior. Second, instead of financial provision or daily care, these men define a responsible father as someone who: acknowledges paternity to the child, mother, and his local community; spends sufficient time with the child to be at least a mentor or "Big Brother" figure; monitors the child's home; meets the child's basic financial needs before spending money on luxuries for himself; minimizes absences in the child's life; and voluntarily distances himself from the child when it is in the child's best interest. I analyze these findings in light of the common definition of responsible fatherhood and suggest several possible theoretical explanations to explain the divergence from this definition. Third, I find that low-income men experience surveillance through three state institutions: child support enforcement, the criminal justice system, and child protective services. They resisted this surveillance primarily by becoming invisible and dropping "off the radar." Men justified their resistance in five ways: they had their own material needs, they did not want the child, they did not want to separate from their child's mother, compliance was unnecessary, or they were incompetent to comply. I analyze these findings in light of Foucault's theory of state social control which contrasts state responses to leprosy and the plague.

Concurrent systems, including asynchronous circuits, computer networks, and multi-threaded programs, have important applications, but they are also very complex and expensive to test. This thesis studies how to test concurrent systems through contexts consisting of queues. Queues, modeling buffers and communication delays, are an integral part of the test settings for concurrent systems. However, queues can also distort the behavior of the concurrent system as observed by the tester, so one should take into account the queues when defining conformance relations or deriving tests. On the other hand, queues can cause state explosion, so one should avoid testing them if they are reliable or have already been tested. To solve these problems, we propose two different solutions. The first solution is to derive tests using some test selection criteria such as test purposes, fault coverage, and transition coverage. The second solution is to compensate for the problems caused by the queues so that testers do not discern the presence of the queues in the first place. Unifying the presentation of the two solutions, we consider in a general testing framework partial specifications, various contexts, and a hierarchy of conformance relations. Case studies on test derivation for asynchronous circuits, communication protocols, and multi-threaded programs are presented to demonstrate the applications of the results.

No description available.

The Internet has experienced a tremendous growth over the past two decades and with that growth have come severe congestion problems. Research efforts to alleviate the congestion problem can broadly be classified into three groups: Cl) Router based congestion detection; (2) Generation and transmission of congestion notification signal to the traffic sources; (3) End-to-end algorithms which control the flow of traffic between the end hosts. This dissertation has largely addressed the first two groups which are basically router initiated. Router based congestion detection mechanisms, commonly known as Active Queue Management (AQM), can be classified into two groups: conventional mathematical analytical techniques and fuzzy logic based techniques. Research has shown that fuzzy logic techniques are more effective and robust compared to the conventional techniques because they do not rely on the availability of a precise mathematical model of Internet. They use linguistic knowledge and are, therefore, better placed to handle the complexities associated with the non-linearity and dynamics of the Internet. In spite of all these developments, there still exists ample room for improvement because, practically, there has been a slow deployment of AQM mechanisms. In the first part of this dissertation, we study the major AQM schemes in both the conventional and the fuzzy logic domain in order to uncover the problems that have hampered their deployment in practical implementations. Based on the findings from this study, we model the Internet congestion problem as a multi-objective problem. We propose a Fuzzy Logic Congestion Detection (FLCD) which synergistically combines the good characteristics of the fuzzy approaches with those of the conventional approaches. We design the membership functions (MFs) of the FLCD algorithm automatically by using Multi-objective Particle Swarm Optimization (MOPSO), a population based stochastic optimization algorithm. This enables the FLCD algorithm to achieve optimal performance on all the major objectives of Internet congestion control. The FLCD algorithm is compared with the basic Fuzzy Logic AQM and the Random Explicit Marking (REM) algorithms on a best effort network. Simulation results show that the FLCD algorithm provides high link utilization whilst maintaining lower jitter and packet loss. It also exhibits higher fairness and stability compared to its basic variant and REM. We extend this concept to Proportional Differentiated Services network environment where the FLCD algorithm outperforms the traditional Weighted RED algorithm. We also propose self learning and organization structures which enable the FLCD algorithm to achieve a more stable queue, lower packet losses and UDP traffic delay in dynamic traffic environments on both wired and wireless networks. In the second part of this dissertation, we present the congestion notification mechanisms which have been proposed for wired and satellite networks. We propose an FLCD based dual explicit congestion notification algorithm which combines the merits of the Explicit Congestion Notification (ECN) and the Backward Explicit Congestion Notification (BECN) mechanisms. In this proposal, the ECN mechanism is invoked based on the packet marking probability while the BECN mechanism is invoked based on the BECN parameter which helps to ensure that BECN is invoked only when congestion is severe. Motivated by the fact that TCP reacts to tbe congestion notification signal only once during a round trip time (RTT), we propose an RTT based BECN decay function. This reduces the invocation of the BECN mechanism and resultantly the generation of reverse traffic during an RTT. Compared to the traditional explicit notification mechanisms, simulation results show that the new approach exhibits lower packet loss rates and higher queue stability on wired networks. It also exhibits lower packet loss rates, higher good-put and link utilization on satellite networks. We also observe that the BECN decay function reduces reverse traffic significantly on both wired and satellite networks while ensuring that performance remains virtually the same as in the algorithm without BECN traffic reduction. / Thesis (M.Sc.Eng.)-University of KwaZulu-Natal, 2006.

The Internet has grown enormously since the introduction of the World Wide Web in the early 90's. The evolution and wide spread deployment of cellular networks have contributed to make the Internet accessible to more people in more places. The cellular networks of today offer data rates high enough for most Internet services. Even so, the service quality experienced by the users is often lower than in wired networks. The performance of TCP has a large impact on user experience. Therefore, we investigate TCP in cellular networks and propose functionality to improve the situation for TCP. We have studied sources of delay and data loss, such as link layer retransmissions, queuing, and handover. Measurements were conducted in a GSM/GPRS testbed. The results indicate that TCP interact efficiently with the GSM link layer protocol in most cases. From experiments of queuing in GPRS, we conclude that with a smaller buffer delay is reduced significantly, but that TCP throughput is about the same as with a larger buffer. Furthermore, we propose an improved buffer management when a connection loses all its resources to traffic with higher priority. We also propose a scheme for data forwarding to avoid negative impact on TCP during handover for WINNER, a research system that was used to test ideas for LTE. The achievable data rates in cellular networks are limited by inter-cell interference that vary over the cell. Inter-cell interference can be mitigated with Coordinated Multipoint techniques (CoMP), techniques that currently are being standardized for LTE-Advanced. System wide CoMP is, however, not an option, since it would be too resource consuming. In order to limit the required resources for CoMP, we propose an approach to select a subset of users for CoMP that is based on user experience. Simulation results indicate that user experience, represented with application utility,  and fairness are improved compared to if only rate is considered in the user selection. / The Internet has grown enormously since the introduction of the World Wide Web in the early 90's. The evolution and wide spread deployment of cellular networks have contributed to make the Internet accessible to more people in more places. The cellular networks of today offer data rates high enough for most Internet services. Even so, the service quality experienced by the users is often lower than in wired networks. The performance of TCP has a large impact on user experience. Therefore, we investigate TCP in cellular networks and propose functionality to improve the situation for TCP. We have studied sources of delay and data loss, such as link layer retransmissions, queuing, and handover. The achievable data rates in cellular networks are limited by inter-cell interference that vary over the cell area. Inter-cell interference can be mitigated with Coordinated Multipoint techniques (CoMP), techniques that currently are being standardized for LTE-Advanced. System wide CoMP is, however, not an option, since it would be too resource consuming. In order to limit the required resources for CoMP, we propose an approach to select a subset of the users for CoMP that is based on user experience.

Call centers have been introduced with great success by many service‐oriented companies. They become the main point of contact with the customer, and an integral part of the majority of corporations. The large‐scale emergence of call centers has created a fertile source of management issues. In this PhD thesis, we focus on various operations management issues of multi‐skill and multichannel call centers. The objective of our work is to derive, both qualitative and quantitative, results for practical management. In the first part, we focus on architectures with limited flexibility for multi‐skill call centers. The context is that of call centers with asymmetric parameters: unbalanced workload, different service requirements, a predominant customer type, unbalanced abandonments and high costs of crosstraining. The most knowing architectures with limited flexibility such as chaining fail against such asymmetry. We propose a new architecture referred to as single pooling with only two skills per agent and we demonstrate its efficiency under various situations of asymmetry. In the second part, we focus on routing problems in multi‐channel call centers. In the first study, we consider a blended call center with calls arriving over time and an infinitely backlogged queue of emails. The call service is characterized by three successive stages where the second one is a break. We focus on the optimization of the email routing to agents. The objective is to maximize the throughput of emails subject to a constraint on the call waiting time. Various guidelines to call center managers are provided. In particular, we prove for the optimal routing that all the time at least one of the two email routing parameters has an extreme value. In the second study, we examine a threshold policy on the reservation of agents for the inbound calls. We study a general non‐stationary model where the call arrival follows a non‐homogeneous Poisson process. The optimization problem consists on maximizing the throughput of outbound tasks under a constraint on the waiting time of inbound calls. We propose an efficient adaptive threshold policy easy to implement. This scheduling policy is evaluated through a comparison with the optimal performance measures found in the case of a constant stationary arrival rate, and also a comparison with other intuitive adaptive threshold policies in the general non‐stationary case. In the third study, we consider a call center model with a call back option, which allows to transform an inbound call into an outbound one. The optimization problem consists on minimizing the expected waiting time of the outbound calls while respecting a service level constraint on the inbound ones. We propose a routing policy with two thresholds, one on the reservation of the agents for inbound calls, and another on the number of waiting outbound calls. A curve relating the two thresholds is determined.