Engineering (Faculty of)
http://hdl.handle.net/10012/9900
Thu, 14 Nov 2019 06:27:27 GMT2019-11-14T06:27:27ZOn Occupancy Based Randomized Load Balancing for Large Systems with General Distributions
http://hdl.handle.net/10012/15250
On Occupancy Based Randomized Load Balancing for Large Systems with General Distributions
Vasantam, Thirupathaiah
Multi-server architectures are ubiquitous in today's information infrastructure whether for supporting cloud services, web servers, or for distributed storage. The performance of multi-server systems is highly dependent on the load distribution. This is affected by the use of load balancing strategies. Since both latency and blocking are important features, it is most reasonable to route an incoming job to a server that is lightly loaded. Hence a good load balancing policy should be dependent on the states of servers. Since obtaining information about the remaining workload of servers for every arrival is very hard, it is preferable to design load balancing policies that depend on occupancy or the number of progressing jobs of servers. Furthermore, if the system has a large number of servers, it is not practical to use the occupancy information of all the servers to dispatch or route an arrival due to high communication cost. In large-scale systems that have tens of thousands of servers, the policies which use the occupancy information of only a finite number of randomly selected servers to dispatch an arrival result in lower implementation cost than the policies which use the occupancy information of all the servers. Such policies are referred to as occupancy based randomized load balancing policies.
Motivated by cloud computing systems and web-server farms, we study two types of models. In the first model, each server is an Erlang loss server, and this model is an abstraction of Infrastructure-as-a-Service (IaaS) clouds. The second model we consider is one with processor sharing servers that is an abstraction of web-server farms which serve requests in a round-robin manner with small time granularity. The performance criterion for web-servers is the response time or the latency for the request to be processed. In most prior works, the analysis of these models was restricted to the case of exponential job length distributions and in this dissertation we study the case of general job length distributions.
To analyze the impact of a load balancing policy, we need to develop models for the system's dynamics. In this dissertation, we show that one can construct useful Markovian models. For occupancy based randomized routing policies, due to complex inter-dependencies between servers, an exact analysis is mostly intractable. However, we show that the multi-server systems that have an occupancy based randomized load balancing policy are examples of weakly interacting particle systems. In these systems, servers are interacting particles whose states lie in an uncountable state space. We develop a mean-field analysis to understand a server's behavior as the number of servers becomes large. We show that under certain assumptions, as the number of servers increases, the sequence of empirical measure-valued Markov processes which model the systems' dynamics converges to a deterministic measure-valued process referred to as the mean-field limit. We observe that the mean-field equations correspond to the dynamics of the distribution of a non-linear Markov process. A consequence of having the mean-field limit is that under minor and natural assumptions on the initial states of servers, any finite set of servers can be shown to be independent of each other as the number of servers goes to infinity. Furthermore, the mean-field limit approximates each server's distribution in the transient regime when the number of servers is large.
A salient feature of loss and processor sharing systems in the setting where their time evolution can be modeled by reversible Markov processes is that their stationary occupancy distribution is insensitive to the type of job length distribution; it depends only on the average job length but not on the type of the distribution. This property does not hold when the number of servers is finite in our context due to lack of reversibility. We show however that the fixed-point of the mean-field is insensitive to the job length distributions for all occupancy based randomized load balancing policies when the fixed-point is unique for job lengths that have exponential distributions. We also provide some deeper insights into the relationship between the mean-field and the distributions of servers and the empirical measure in the stationary regime.
Finally, we address the accuracy of mean-field approximations in the case of loss models. To do so we establish a functional central limit theorem under the assumption that the job lengths have exponential distributions. We show that a suitably scaled fluctuation of the stochastic empirical process around the mean-field converges to an Ornstein-Uhlenbeck process. Our analysis is also valid for the Halfin-Whitt regime in which servers are critically loaded. We then exploit the functional central limit theorem to quantify the error between the actual blocking probability of the system with a large number of servers and the blocking probability obtained from the fixed-point of the mean-field. In the Halfin-Whitt regime, the error is of the order inverse square root of the number of servers. On the other hand, for a light load regime, the error is smaller than the inverse square root of the number of servers.
Wed, 13 Nov 2019 00:00:00 GMThttp://hdl.handle.net/10012/152502019-11-13T00:00:00ZScalable Surfaces for Electromagnetic Energy Harvesting and Wireless Power Transfer
http://hdl.handle.net/10012/15244
Scalable Surfaces for Electromagnetic Energy Harvesting and Wireless Power Transfer
Erkmen, Faruk
The idea of collecting electromagnetic (EM) energy and converting it into various forms of useful power dates back to the early 20th century. Nikola Tesla's wireless power transfer experiments demonstrated the concept first, which was followed by researchers in Japan and the USA in subsequent decades. In terms of a working prototype, the first rectenna for efficient reception and rectification of microwave power was developed in the early 1960s. Later, the introduction of semiconductor diodes and the invention of Schottky diodes were significant developments towards the realization of practical rectennas. Since then, owing to the numerous applications in different technology domains (i.e. consumer electronics, renewable energy, transportation, internet of things, artificial intelligence, telecommunications, defense & space, biomedical engineering), wireless power transfer and EM energy harvesting have attracted significant interest.
Harvesting the ambient EM energy has emerged more recently as a promising application with potential for commercial success and contribution to a sustainable future with renewable energy. Many studies have reported the available ambient power densities measured in several parts of the world demonstrating the potentials and limitations of the concept. Traditional single rectenna structures have found very little use due to their inherent limitations at low power densities. Large rectenna arrays or periodic structures covering larger surface areas have become particularly important in order to efficiently harvest and convert the energy.
A rectenna consists of two main functional building blocks: the rectifier and the EM collector. The work in this thesis first focuses on improving these functional blocks individually. Regarding the rectifier function; a balanced full-wave rectifier is proposed where the circuit is differentially fed by two separate antennas. This configuration allows the received power to be rectified and transferred into a load between two antennas, making it convenient to channel the harvested power in rectenna arrays. The proposed concept is demonstrated using an array of T-matched dipole antennas at 2.45 GHz. It is also compared with half-wave rectennas that occupy the same footprint with an identical array layout. Measurement results show that, under the same circumstances, the proposed full-wave rectification performs better than the traditional half-wave rectification and it is indeed suitable for energy harvesting rectenna arrays.
Regarding the EM collector; a novel Frequency Selective Surface (FSS) is developed as an absorber surface that accepts 98.5% of the available power and collects 97% of it exclusively on its resistive load (only 1.5% is dissipated as dielectric and metallic losses). To demonstrate its performance, a proof of concept FSS absorber is fabricated and its resistive load is replaced with a matched full-wave rectifier. Measurement results show that the overall Radiation-to-dc conversion efficiency of the complete rectenna system reaches 61%, which is considerably higher than the previously reported FSS based rectennas.
Subsequent sections in this thesis expand the energy harvesting surface by adding dual-band and dual-polarization capabilities. Design details and simulation results are provided together with measurement results. Fabricated prototypes are tested and their overall performance is evaluated based on the rectified DC power at the system load as percentage of the available EM power on the physical surface area of the rectenna (i.e. radiation-to-dc conversion efficiency).
A key contribution of this thesis is the introduction of the scalability concept for energy harvesting. The periodic absorber surfaces presented in this thesis have built-in channelling features that allow multi-cell configurations to feed a single rectifier. This is demonstrated to be an efficient means to increase the EM collector area per rectifier by effortlessly scaling the surface area while efficiently channelling the collected power. As a result, the number of diodes and diode losses are minimized in the system, leading to higher overall rectenna efficiencies. Real life ambient power densities can be on the order of nW/cm2 and the work in this thesis show that larger EM collectors can significantly mitigate the limitations posed by such low power levels. As an example; when integrated with a multi-cell configuration, an ordinary rectifier made with Schottky diodes was efficiently used at a power density that is less than 1/12th of that would be required if the same rectifier were to be used with a traditional single unit cell approach.
Thu, 31 Oct 2019 00:00:00 GMThttp://hdl.handle.net/10012/152442019-10-31T00:00:00ZEvaporation of Liquid Wall Film in Direct Injection Spark Ignition Engine-like Conditions
http://hdl.handle.net/10012/15239
Evaporation of Liquid Wall Film in Direct Injection Spark Ignition Engine-like Conditions
Yang, Song
The liquid wall film formed by the spray impingement in Direct Injection Spark Ignition (DISI) engines can directly produce a large amount of Particle Matter (PM) emissions. The PM emissions can be tremendously reduced if all the liquid wall film can evaporate completely before flame propagates to the wall surface and the combustion of ‘pool fire’ fed by the evaporating liquid wall film can be totally eliminated. Evaporation models are widely used to predict the evaporation of liquid wall film in engines, but requiring accurate mass transfer correlations. However, it is challenging to experimentally determine the accurate mass transfer correlations of the liquid wall film in engines; since the evaporation time of the thin liquid wall film in engines is quite short and the thickness of the liquid wall film is extremely thin. Thus, numerical simulation has become a useful tool to provide insight into the underlying transient evaporation characteristics of liquid wall film in DISI engine-like conditions and to derive the mass transfer correlations.
In this thesis research, numerical study has been conducted for a two-dimensional, two-phase, transient, non-isothermal and species transport problem representing the evaporation of liquid wall film in DISI engine-like conditions. The unique features of the numerical models are the inclusion of the transient motion and heating of the liquid phase, the blowing effects caused by evaporation, and the variation of thermo-physical properties. The governing equations which mathematically describe the transient evaporation process of liquid wall film in DISI engines, are discretized and solved using a Finite Volume Method (FVM) based software, Fluent, with its capability of User Defined Function (UDF) programming.
The numerical evaporation models are validated with existing analytical and experimental data; and good agreements are observed. Subsequently, the validated models are used for the numerical study of the evaporating liquid wall film in DISI engine-like conditions to investigate its transient evaporation characteristics and determine its mass transfer correlations. The results show that the evaporation rate of liquid wall film, characterized by mass transfer coefficient, is non-uniform along the wall film, which is consistent with the development of species boundary layer and the decline of species concentration gradient within the boundary layer. The transient evaporation of liquid wall film in DISI engine-like conditions is mainly determined by the gas/liquid interfacial temperature, which can be directly affected by the transient heating of the liquid phase. The newly developed mass transfer correlations taking into account the blowing effects and effects caused by convection and the variation of thermo-physical properties during the transient evaporation process of the liquid wall film can predict their evaporation rate much more accurately than the existing correlations available in literature.
Wed, 30 Oct 2019 00:00:00 GMThttp://hdl.handle.net/10012/152392019-10-30T00:00:00ZCoupled Dynamics of Cable-Harnessed Structures: Experimental Validation
http://hdl.handle.net/10012/15236
Coupled Dynamics of Cable-Harnessed Structures: Experimental Validation
Yerrapragada, Karthik; Salehian, Armaghan
The experimental study and model validations for the coupled dynamics of a cable-harnessed beam structure are presented. The system under consideration consists of multiple pretensioned cables attached along the length of the host beam structure positioned at an offset distance from the beam centerline. Analytical model presented by the coupled partial differential equations (PDEs) for various coordinates of vibrations is found, and the displacement frequency response functions (FRFs) obtained for both Euler–Bernoulli and Timoshenko-based models are compared to those from the experiments for validation. The results are shown to be in very good agreement with the experiments.
Mon, 15 Jul 2019 00:00:00 GMThttp://hdl.handle.net/10012/152362019-07-15T00:00:00Z