Title Information
Title
C-MR: Continuous Execution of MapReduce Workflows for Stream Processing
Name: Personal
Name Part
Backman, Nathan John
Role
Role Term: Text
creator
Origin Information
Copyright Date
2013
Physical Description
Extent
x, 98 p.
digitalOrigin
born digital
Note
Thesis (Ph.D. -- Brown University (2013)
Name: Personal
Name Part
Cetintemel, Ugur
Role
Role Term: Text
Director
Name: Personal
Name Part
Fonseca, Rodrigo
Role
Role Term: Text
Reader
Name: Personal
Name Part
Zdonik, Stanley
Role
Role Term: Text
Reader
Name: Corporate
Name Part
Brown University. Computer Science
Role
Role Term: Text
sponsor
Genre (aat)
theses
Subject
Topic
Stream processing
Subject
Topic
distributed computing
Subject (FAST) (authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/1134622")
Topic
Streamflow--Data processing
Subject (FAST) (authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/906987")
Topic
Electronic data processing--Distributed processing
Record Information
Record Content Source (marcorg)
RPB
Record Creation Date (encoding="iso8601")
20131217
Language
Language Term: Code (ISO639-2B)
eng
Language Term: Text
English
Abstract
Data processing frameworks provide application programmers an interface to manipulate and analyze data. This thesis studies a novel parallel stream processing model, designed for workflow-based data processing frameworks, that leverages application performance requirements to motivate the flexible scheduling and fine-grained allocation of data to computing nodes.<br/> <br/> We feature this processing model through the design and implementation of the Continuous-MapReduce (C-MR) data processing framework. C-MR abstracts away the complexities of parallel stream processing and workflow scheduling while providing the simple and familiar MapReduce programming interface with the addition of stream window semantics. Its novel processing model enables: 1) fine-grained, workflow-wide load balancing across computing nodes; 2) the evolving application of data and task parallelism models as guided by application performance requirements; and 3) a novel scheduling framework which supports gradual transitions between scheduling policies relative to application performance and/or resource availability.<br/> <br/> This work explores the potential of the C-MR processing model by studying our single-host implementation of C-MR that supports parallel execution on non-dedicated and heterogeneous computing nodes (both multi-core CPUs and GPUs). We then study this processing model through the implementation of a distributed version of C-MR that supports execution on multiple hosts. This endeavor involved the generalizable strategy of employing hierarchical instances of the C-MR processing model while requiring modifications to the data acquisition and load balancing strategies. Experimental results from these studies show that the C-MR processing model can effectively support the continuous execution of workflows of MapReduce jobs for stream processing while being resilient to stream and resource fluctuations due to the processing model's flexibility and diversification of processing responsibilities.<br/>
Identifier: DOI
10.7301/Z0JD4V3C
Access Condition: rights statement (href="http://rightsstatements.org/vocab/InC/1.0/")
In Copyright
Access Condition: restriction on access
Collection is open for research.
Type of Resource (primo)
dissertations