Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


osu1126208308.pdf (1.45 MB)

ETD Abstract Container

Abstract Header

Advanced middleware support for distributed data-intensive applications

Du, Wei
http://rave.ohiolink.edu/etdc/view?acc_num=osu1126208308

Abstract Details

2005, Doctor of Philosophy, Ohio State University, Computer and Information Science.
In recent years, the size of datasets has been growing dramatically, and these datasets are often distributed across multiple sites. Processing such large-scale distributed datasets plays an increasingly important role in many domains. The broad goal of our work is to provide advanced middleware support to ease the development of distributed data-intensive applications, and facilitate their efficient execution. A coarse-grained pipelined execution model provides a natural vehicle for executing such applications. Here, the processing associated with an application is carried out at several stages, which are executed on a pipeline of computing units. Typically, the first stage in this pipeline is the unit where the input data is available, and the last stage is where the final results are to be viewed. In this thesis, I present language and compiler supports for developing distributed data-intensive applications in a coarse-grained pipelined fashion. These supports could allow programmers to focus on writing a sequential code to specify the application-specific tasks. Our extended language constructs can help expose both pipelined and data parallelism to the compiler. Then our compilation system is responsible for selecting a set of candidate filter boundaries, determining the volume of communication required if a particular boundary is chosen, performing the decomposition, and generating code for execution on coarse-grained pipelined execution model. To support the adaptivity aspect of the applications, we adopted a hybrid methodology, which combines compile-time analysis with runtime feedback. A program analysis algorithm states the execution time of an application component as a function of the values of the adaptation parameters and other runtime constants. These constants are determined by initial runs of the application in the target environment. Based on these calculated constants, adaptation parameters can be modified and corresponding adjustment steps allow user to achieve a given change in execution time, or to retain the same performance under a particular variation in available resources. Substantial amount of experiments have been carried out to assess this dissertation, which shows the proposed algorithms and models work quite effectively in practice. The dissertation concludes with a set of open research questions that frame the future work.
Gagan Agrawal (Advisor)
203 p.
Computer Science
Distributed Data-intensive Applications; Coarse-Grained Pipelined Parallelism; Language and Compiler Support; Program Decomposition; Communication Granularity Selection; Program Adaptation

Recommended Citations

Citations

  • Du, W. (2005). Advanced middleware support for distributed data-intensive applications [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1126208308

    APA Style (7th edition)

  • Du, Wei. Advanced middleware support for distributed data-intensive applications. 2005. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1126208308.

    MLA Style (8th edition)

  • Du, Wei. "Advanced middleware support for distributed data-intensive applications." Doctoral dissertation, Ohio State University, 2005. http://rave.ohiolink.edu/etdc/view?acc_num=osu1126208308

    Chicago Manual of Style (17th edition)

osu1126208308
1,216
© 2005, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.