No description available.

No description available.

Recently, Electric Vehicles (EVs) have received extensive consideration since they offer a more sustainable and greener transportation alternative compared to fossil-fuel propelled vehicles. Lithium-ion batteries are increasingly being considered in EVs due to their high energy density, slow loss of charge when not in use, and for lack of hysteresis effect. Conventionally, the batteries are connected in series to achieve the load voltage requirements. However, for the batteries with intrinsic discrepancies or different initial states, cell balancing is a concern because it is the weakest cell that determines the empty point for the battery and an undercharged series cell will shorten the lifetime of the entire pack. The imbalance potential of the battery behaves as the way of State-of-Charge (SOC) mismatch and it’s also temperature dependent. Therefore, in this thesis, an active cell balancing optimization was proposed and conducted in MATLAB to optimize battery unused capacity and thermal effect simultaneously based on bi-directional balancing system and pre-estimated SOC. The bi-directional balancing system was physically built based on “Fly-back” converter to compare balancing performance in discharging, idle, and plug-in charging mode. Moreover, a battery combined model worked collaboratively with robust state and parameter estimation strategies, namely Extended Kalman Filter (EKF) and Smooth Variable Structure Filter (SVSF) in order to estimate SOC for cell balancing. As a result, the proposed method can effectively optimize SOC mismatch around 2.5%. Meanwhile, more uniform temperature was achieved and the maximum temperature can be reduced about 7 ℃. / Thesis / Master of Applied Science (MASc)

No / Over the last few years, cloud computing has become quite popular. It offers Web-based companies the advantage of scalability. However, this scalability adds complexity which makes analysis and predictable performance difficult. There is a growing body of research on load balancing in cloud data centres which studies the problem from the perspective of the cloud provider. Nevertheless, the load balancing of scalable web servers deployed on the cloud has been subjected to less research. This paper introduces a simple queueing model to analyse the performance metrics of web server under varying traffic loads. This assists web server managers to manage their clusters and understand the trade-off between QoS and cost. In this proposed model two thresholds are used to control the scaling process. A discrete-event simulation (DES) is presented and validated via an analytical solution.

The SiC MOSFET has attracted interest due to its superior characteristics compared to its Si counterpart. Several SiC MOSFETs are usually paralleled to increase current capability, considering cost effectiveness and manufacturability. Current unbalance among the MOSFETs is a concern as it affects reliability. The two main causes are asymmetrical layout and parameter mismatch. The variation in parameters, unlike circuit or module layout, is unavoidable during production. Among all the parameters of MOSFET, the spreads in on-state resistance (Rds(on)) and threshold voltage (Vth) are the major concerns during paralleling. The disparity in Rds(on) causes static current unbalance which is self-limited due to the positive temperature coefficient of Rds(on). Its influence is not investigated here. The threshold voltage Vth has a negative temperature coefficient, forcing the MOSFET with lower Vth to carry more current during switching transient. Paralleled MOSFETs are usually de-rated to guarantee safe operation. Balancing of peak currents during switching transient isthe goal of this work. Integration of current/voltage sensors into paralleled structure is difficult in real application. Complicated feedback loop design and separate gate drivers also need to be avoided in perspective of cost and volume. Passive balancing solutions are investigated in this dissertation. The inductors and resistors most effective in improving current sharing are identified by parametric analysis. Their current balancing mechanisms are analyzed in circuit point of view. The design guidelines involving the magnitude of Vth mismatch, current rise time, and unbalance percentage are derived for the selection of passive components. The theory upholds well when substantial parasitics from device package and layout exist. Several passive balancing structures are analyzed and compared in terms of current balancing capability, voltage stress, total switching loss, and switching loss difference. All of them can provide much better current and power balancing without increasing switching loss. Some of the them may increase the stress-inducing inductance, which can be reduced by negative magnetic coupling. Perfect coupling between power-source inductors would enable current matching without penalty on voltage stress. Common-source inductance (Lcm) is effective in dynamic balancing, but at the expense of higher switching loss. It is not considered in power module application because Kelvin connection is normally applied. However, wire bond inside the package of discrete MOSFETs and part of the external leads are inevitable and add to Lcm. Peak-current and switching energy mismatches vary with operating conditions (including input voltage, input current, and switching speed). Design guidelines and procedures that are valid for wide operating range are provided for cases with and without Lcm. This dissertation also models the switching energy and switching energy mismatch of paralleled MOSFETs. The influence of operating conditions, passive balancing components, layout and package parasitic inductances, nonlinear channel performance, and voltage dependent parasitic capacitors are included in the modeling process. The resulting high order system is simplified by reducing the number of passive components and number of devices without losing accuracy. The influence of current balancing components and magnitude of threshold voltage mismatch on sharing are discussed based on modeling results. In conclusion, this dissertation balances the transient currents between paralleled SiC MOSFETs automatically by inductance, resistance and magnetic coupling. This procedure is done utilizing one gate driver without current/voltage sensors and feedback loop. Those solutions work for both polarities of Vth mismatch and force balancing from the first current peak. Design guidelines involving the magnitude of Vth mismatch, current rise time, and maximum peak-current difference are derived to guide the choice of passive components. The detail design procedures are recommended to force currents to share over wide operating range. The aforementioned benefits are demonstrated by two paralleled SiC MOSFETs (C2M0160120D) tested at variant operating conditions. The difference of peak currents can be reduced below 5% of steady-state current in every switching transient. Switching energy mismatch percentage can be reduced by 6 times without increasing total switching energy. / Ph. D.

NAND flash-based Solid State Devices (SSDs) provide high performance and energy efficiency and at the same time their capacity continues to grow at an unprecedented rate. As a result, SSDs are increasingly being used in high end computing systems such as supercomputing clusters. However, one of the biggest impediments to large scale deployments is the limited erase cycles in flash devices. The natural skewness in I/O workloads can results in Wear imbalance which has a significant impact on the reliability, performance as well as lifetime of the cluster. Current load balancers for storage systems are designed with a critical goal to optimize performance. Data migration techniques are used to handle wear balancing but they suffer from a huge metadata overhead and extra erasures. To overcome these problems, we propose an endurance-aware write off-loading technique (EWO) for balancing the wear across different flash-based servers with minimal extra cost. Extant wear leveling algorithms are designed for a single flash device. With the use of flash devices in enterprise server storage, the wear leveling algorithms need to take into account the variance of the wear at the cluster level. EWO exploits the out-of-place update feature of flash memory by off- loading the writes across flash servers instead of moving data across flash servers to mitigate extra-wear cost. To evenly distribute erasures to flash servers, EWO off-loads writes from the flash servers with high erase cycles to the ones with low erase cycles by first quantitatively calculating the amount of writes based on the frequency of garbage collection. To reduce metadata overhead caused by write off-loading, EWO employs a hot-slice off-loading policy to explore the trade-offs between extra-wear cost and metadata overhead. Evaluation on a 50 to 200 node SSD cluster shows that EWO outperforms data migration based wear balancing techniques, reducing up to 70% aggregate extra erase cycles while improving the write performance by up to 20% compared to data migration. / Master of Science

A methodology for the solution of single-model, stochastic assembly line balancing problem is developed for the objective of minimizing the total labor cost (dictated by the number of stations on the line) and expected incompletion cost arising from tasks not completed within the prespecified cycle time. The proposed procedure is an approximation procedure that divides the problem into subproblems. For each subproblem, an approximate solution is obtained using the dynamic programming procedure developed for the problem. This procedure is incorporated with a special bounding strategy to overcome the rapidly increasing storage and computational requirements as the size of the problem increases. These approximate solutions are further improved by a branch-and-bound type of procedure called the improvement procedure. This procedure uses approximate costs, instead of lower bounds, to fathom the nodes of the enumeration tree constructed; thus, it is not, in the true sense of the word, the branch-and-bound technique. Consequently, the procedure is not guaranteed to result in the optimal solution; however, it is shown to generate solutions within (1 + ε) of the optimal solution. The improvement procedure either improves the approximate solutions obtained using the dynamic programming procedure or determines that they are quite close to the optimal ones. The improved solutions of the subproblems are then appended to each other to produce the solution of the original problem. Some dominance properties that contribute to the effectiveness of the improvement procedure and help in reducing the size of the enumeration tree are developed. Some sequencing and scheduling problems related to the node evaluation scheme of the improvement procedure are also investigated. A single-machine sequencing procedure is developed for the objective of minimizing the expected incompletion cost with tasks having a common due date and stochastic processing times. This procedure is extended to construct a schedule on M parallel machines. In these procedures, in-completion costs of the tasks are independent of their expected performance times; it can be interpreted as relaxing the precedence relations among the tasks. Solution procedures are also developed for the above sequencing and scheduling problems for the case in which the incompletion costs of the tasks are proportional to their expected performance times. Computational results and analyses made indicate that these procedures result in almost optimal solutions. / Ph. D.

Approved for public release; distribution in unlimited. / This thesis develops and implements a Control Moment Gyroscope (CMG) steering law, controller and active balancing system for a three-axis satellite simulator (TASS). The CMGs are configured in a typical pyramid configuration (the fourth CMG position being null). The development was done primarily with simulation and experiments utilizing Real Time Workshop and XPC Target of MATLAB and SIMULINK. The TASS is a double circular platform mounted on a spherical air bearing with the center of rotation (CR) about the approximate physical geometric center of the simulator. The TASS utilizes three moveable masses in the three body axes for balancing which actively eliminate any center of gravity (CG) offset and return the CG to the CR. The TASS supports an optics payload designed to acquire, track and point a received laser beam onto an off-satellite target. The target may be stationary or moving. Actively balancing the TASS reduces the torque output requirement for the CMGs while maintaining either a stabilized level platform or a particular commanded attitude. Reduction or elimination of torque output from the CMGs results in a more stabilized platform, less structural induced vibration, less jitter in payload optics and less power required in spacecraft applications. / Lieutenant Commander, United States Navy

We developed a diffusive load balancing scheme that equalizes the available capacities of nodes in a peer-to-peer (P2P) system. These nodes may have different resource capacities, geographic locations, or availabilities (i.e., length of time being part of the peer-to-peer system). The services on these nodes may have different service times and arrival rates of requests. Using the diffusive scheme, the system is able to maintain similar response times for its services. Our scheme is a modification of the diffusive load balancing algorithms proposed for parallel computing systems. This scheme is able to handle services with heterogeneous resource requirements and P2P nodes with heterogeneous capacities. We also adapted the diffusive scheme to clustered peer-to-peer system, where a load balancing operation may move services or nodes between clusters. After a literature survey of this field, this thesis investigates the following issues using analytical reasoning and extensive simulation studies. The load balancing operations equalize the available capacities of the nodes in a neighborhood to their averages. As a result, the available capacities of all nodes in the P2P system converge to a global average. We found that this convergence is faster when the scheme uses neighborhoods defined by the structure of the structured P2P overlay network rather than using randomly selected neighbors. For a system with churn (i.e. nodes joining and leaving), the load balancing operations maintain the standard deviation of the available capacities of nodes within a bound. This bound depends on the amount of churn and the frequency of load balancing operations, as well as on the capacities of the nodes. However, the sizes of the services have little impact on this bound. In a clustered peer-to-peer system, the size of the bound largely depends on the average cluster size. When nodes are moved among clusters for load balancing, the numbers of cluster splits and merges are reduced. This may reduce the maintenance cost of the overlay network.

We developed a diffusive load balancing scheme that equalizes the available capacities of nodes in a peer-to-peer (P2P) system. These nodes may have different resource capacities, geographic locations, or availabilities (i.e., length of time being part of the peer-to-peer system). The services on these nodes may have different service times and arrival rates of requests. Using the diffusive scheme, the system is able to maintain similar response times for its services. Our scheme is a modification of the diffusive load balancing algorithms proposed for parallel computing systems. This scheme is able to handle services with heterogeneous resource requirements and P2P nodes with heterogeneous capacities. We also adapted the diffusive scheme to clustered peer-to-peer system, where a load balancing operation may move services or nodes between clusters. After a literature survey of this field, this thesis investigates the following issues using analytical reasoning and extensive simulation studies. The load balancing operations equalize the available capacities of the nodes in a neighborhood to their averages. As a result, the available capacities of all nodes in the P2P system converge to a global average. We found that this convergence is faster when the scheme uses neighborhoods defined by the structure of the structured P2P overlay network rather than using randomly selected neighbors. For a system with churn (i.e. nodes joining and leaving), the load balancing operations maintain the standard deviation of the available capacities of nodes within a bound. This bound depends on the amount of churn and the frequency of load balancing operations, as well as on the capacities of the nodes. However, the sizes of the services have little impact on this bound. In a clustered peer-to-peer system, the size of the bound largely depends on the average cluster size. When nodes are moved among clusters for load balancing, the numbers of cluster splits and merges are reduced. This may reduce the maintenance cost of the overlay network.