Several defenses will be held over the next two weeks. These include Aravind Dwarakacharla on Friday, June 29 at 11 a.m., Zhipeng Ouyang on Monday, July 2 at 1 p.m., and Shivashis Saha on Monday, July 9 at 9:30 a.m. The following are abstracts of the defenses.
Aravind Dwarakacharla
"Template Matching in Comprehensive Two-Dimensional Chromatography Using Multispectral Information"
256C Avery Hall
Abstract:
Comprehensive two-dimensional gas chromatography (GCxGC) is a chromatographic technique that offers superior separation power for complex mixtures of chemical compounds. Mass spectrometry (MS) provides multi-spectral information for the separated chemical constituents produced by GCxGC. Chemical identification is one of the critical tasks in GCxGC-MS analysis. Manual identification of chemical separations is a tedious and time consuming task. Peak pattern matching is a powerful approach that records a pattern of peaks with analytical metadata in a template using prior knowledge of the compounds. Then, the template is matched to target peaks in a new chromatogram and the metadata from the template is copied into the new chromatogram to identify and classify the matched peaks.
This report implements modifications to the peak pattern matching algorithm developed by Mingtian Ni to increase the template matching accuracy. The new approach uses multispectral information rather than or along with the retention-times of the compounds. The peak pattern matching algorithm selects a more appropriate target peak using a spectral similarity measure called match factor as the selection criterion instead of the distance measure between the template peak and target peak. The match factor is calculated by comparing the multispectral signatures of a template peak and target peak. Currently, the peak pattern matching algorithm uses a greedy approach to evaluate one-to-one correspondences between template and target peaks from a set of all possible correspondences. The greedy approach results in a solution, but not always in a correct solution. The Hungarian algorithm is implemented as an alternative to the greedy approach to achieve a better solution. This report evaluates the accuracy of template matching with the proposed changes for GCxGC chromatograms. Experimental results suggest the new approach, with the combination of multispectral information and retention-times, is more accurate and robust.
Zhipeng Ouyang
"On Heterogeneous User Demands In Peer-To-Peer Video Streaming Systems"
347 Avery Hall
Abstract:
Peer-to-Peer (P2P) video streaming systems have been widely deployed in the Internet for distributing video content. A P2P video streaming system usually consists of a large number of peers, which have heterogeneous physical properties, such as various bandwidth capacities, widely-spread geographical locations, and diverse viewing appliances. Orthogonal to the physical heterogeneity, there is another type of heterogeneity called demand heterogeneity. Namely, peers have their own demands on the quality and type of the streaming service. The problem of physical heterogeneity has been extensively studied for P2P video streaming systems. However, the problem of demand heterogeneity has received little attention and as a result current P2P video streaming systems cannot achieve satisfactory performance due to demand heterogeneity. In this dissertation, we study how to design efficient P2P video streaming systems with heterogeneous user demands. Specifically, we consider two types of user demands: first, demands on video quality, and second, demands on playback delay.
First, we study the problem of heterogeneous user demands on video quality. Inspired by the fact that different peers in a P2P live streaming system may watch the same channel with different window sizes which have different quality demands, we design an efficient P2P live streaming system which greatly reduces the bandwidth consumption and still achieves satisfactory streaming quality. In order to reduce the bandwidth consumption, we use adaptive streaming, so that the streaming rate of a peer is commensurate with its window size. In order to maintain satisfactory streaming quality even in the case when peers dynamically change their window sizes, small-window peers are designed to contribute part of their bandwidth to help large- window peers.
Second, we study the problem of heterogeneous user demands on playback delay. Specifically, we consider two types of playback delay demands: time-shifted streaming service and bounded-delay live streaming service. With time-shifted streaming service, a peer can watch a program which has been broadcasted live before the peer joins the system. To provide such a type of streaming service, we propose a cooperation-based prefetching design which distributes video segments more widely in a P2P streaming system and thus greatly improves the streaming quality. With bounded-delay live streaming service, some peers (e.g., paid users) can watch a live broadcast program with short and bounded playback delays, and other peers (e.g., free users) with best-effort short playback delays. We prove that the bounded-delay live streaming problem is NP Complete, and we propose a heuristic algorithm which constructs an efficient P2P overlay structure of peers by considering their playback delay requirements and their physical properties.
The promising results presented in this dissertation show that our proposed algorithms are effective in meeting heterogeneous user demands on video quality and playback delay. In the future, we are interested in designing efficient P2P streaming systems for the large number of mobile users by considering their unique physical heterogeneity and demand heterogeneity.
Shivashis Saha
"Data Center Design: Architecture and Energy Consumption"
256C Avery Hall
Abstract:
Data Centershave evolved from mainframe systems and enterprise networks into sophisticated networks of 100,000 or more servers for supporting next-generation Computing-as-a-Service(CaaS) and Cloudcomputing infrastructures. The evolution of data centers has resulted in an expected exponential increase in both data center network traffic and data center energy consumption. Thus, there is tremendous interest in state-of-the-art research in data centers with the objective of designing efficient data center network(DCN) architectures and developing strategies for minimizing energy consumption.
In this dissertation, we propose HyScaleand SlimNet, cost-effective and scalable DCN architectures using hybrid optical networks. The proposed architectures are fault-tolerant, recursively defined, and have low network complexity. Moreover, the proposed architectures also have several desirable graph-theoretic properties like high bisection-width and low diameter. We propose a non-selfish destination selectionparadigmfor containing the impact of the changes in the availability of the resources at a destination. We also present efficient non-selfishheuristics for minimizing the resource blocking rate.
We adapt energy models for developing comprehensive and state-of-the-art energy models for servers and network elements in data centers. Using the server and network energy models, we propose a greenrouting scheme that minimizes the total combined energy consumption of all servers and all network elements in a data center under dynamic traffic. The proposed green routing scheme uses dynamic voltage scaling, rateadaptation, and anycasttransmission for minimizing the total energy consumption in a data center. Finally, we propose thermal-constrained energy-aware algorithms for partitioning tasks in multi-core multiprocessor systems using both worst-case execution time and actual utilization of the tasks.