Browsing by Subject "Distributed computing"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Data dissemination for distributed computing.
    (2010-02) Kim, Jinoh
    Large-scale distributed systems provide an attractive scalable infrastructure for network applications. However, the loosely-coupled nature of this environment can make data access unpredictable, and in the limit, unavailable. This thesis strives to provide predictability in data access for data-intensive computing in large-scale computational infrastructures. A key requirement for achieving predictability in data access is the ability to estimate network performance for data transfer so that computation tasks can take advantage of the estimation in their deployment or data source selection. This thesis develops a framework called OPEN (Overlay Passive Estimation of Network Performance) for scalable network performance estimation. OPEN provides an estimation of end-to-end accessibility for applications by utilizing past measurements without the use of explicit probing. Unlike existing passive approaches, OPEN is not restricted to pairwise or a single network in utilizing historical information; instead, it shares measurements between nodes without any restrictions. As a result, it achieves n2 estimations by O(n) measurements. In addition, this thesis considers data dissemination in two specific environments. First, we consider a parallel data access environment in which multiple replicated servers can be utilized to download a single data file in parallel. To improve both performance and fault tolerance, we present a new parallel data retrieval algorithm and explore a broad set of resource selection heuristics. Second, we consider collective data access in applications for which group performance is more important than individual performance. In this work, we employ communication makespan as a group performance metric and propose server selection heuristics to maximize collective performance.
  • Thumbnail Image
    Item
    A MPI-based Distributed Computation for Supporting Optimization of Urban Designs with QUIC EnvSim
    (2015-07) Vuggumudi, Viswanadh Kumar Reddy
    In the present day of urbanization, rise in urban infrastructure is causing an increase in air temperatures and pollution concentrations. This leads to an increase in the energy required to cool buildings and more focused efforts to mitigate pollution. An effective way to mitigate these problems is by carefully designing cityscapes i.e., by placing the buildings, vegetation optimally and choosing energy efficient building materials. Researchers have been building computational models to understand the effects of urban infrastructure on microclimate. Simulating these models is a computationally expensive task. QUIC EnvSim (QES) is a dynamic, scalable and high performance framework that has provided a platform for building and simulating these models. QUIC EnvSim uses Graphics Processing Units (GPUs) to run each individual simulation faster than previous simulation codes. Though each individual simulation takes a short time, it is often required to perform large numbers of simulations and it can take a long time to complete them. This thesis introduces MPI QUIC, a scalable and extendable framework for running these simulations across a cluster of machines, effectively reducing the time required to run all simulations. Various tests on the framework have shown that the framework is capable of running large numbers of simulations in a relatively less amount of time. A test running 65536 simulation was performed. The estimated time for running the test on a single computer is approximately 11.37 days, with each simulation taking approximately 15 seconds to complete. The framework was able to finish running all the simulations in 19 hours, 0 minutes and 25 seconds showing a tremendous speed up of 92.5%. Thus urban planners can use this framework along with QUIC EnvSim to understand the effects of urban forms on microclimate and take informed design decision relatively quickly for building environment friendly urban landscapes. Besides providing a distributed computational environment, the other goal of the MPI QUIC project is to provide a user friendly interface for specifying optimization problems. The current work provides the ground work for the successors of the current work to provide a programmable interface for end users for specifying optimization problems. The framework is also designed so that future implementers can incorporate optimization algorithms that can optimize on multiple fitness functions.
  • Thumbnail Image
    Item
    Optimizing urban environmental simulations using boinc
    (2013-08) Vegesna, Aditya
    Urban cities are usually densely populated and have massive infrastructure. They consume a lot of energy and generate pollution. Urban form and structure interact with the environment in a complex way. There is transfer of energy between buildings and the ground layer. Winds flow through the urban street canyons, affecting evaporation, temperature and pollution dispersion. The effects of such complex interactions are still not widely known or understood. How well an urban space disperses pollution, or requires energy for heating or cooling is potentially impacted by many components, such as where the buildings are located with respect to each other, which materials the buildings are constructed from, or where trees or parks are placed. The aim of the Genusis project is to provide a tool for urban planners that they can utilize to understand such impacts and to assist them in taking design decisions accordingly. Even with just a few choices in building locations or tree types the number of possible configurations is vast. Running the simulations on many thousand of these configurations is a huge problem on its own and truly not feasible for urban planners to use in their daily routines.This thesis strives towards tackling that problem by developing a computational environment in which specifying these configurations is easy and can compute potential solutions to the problems within an acceptable time frame using multiple machines. A simple and yet powerful language is created to let urban planners control the simulations and specify the configurations. In order to reduce the computational time, Berkeley Open Infrastructure for Network Computing (BOINC) is used to harness all available computational resources. Experiments were conducted to analyze the implementation and performance of the system. The results obtained validate the implementation and indicate a significant performance gain.