High Availability on a Distributed Real Time Processing System

Enrico Tanuwidjaja

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2015-134
May 15, 2015

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-134.pdf

Our Capstone project involves working with an open source distributed real­ time processing system called Apache Storm, in collaboration with Cisco Systems, Inc. The term “real­ time processing” in this context means that the system is able to respond within seconds or sub-­second to requests, while “distributed” means that it is running on multiple computers. The goal of the project is to add a feature called “k-­safety” to Storm. With k-­safety, Storm will be able to tolerate up to k machine failures without losing data or reducing its response time, making the system highly available. Cisco plans to integrate our modified version of Storm into their data processing pipeline and use it to support internal and customer-­facing products.

Advisor: Michael Franklin


BibTeX citation:

@mastersthesis{Tanuwidjaja:EECS-2015-134,
    Author = {Tanuwidjaja, Enrico},
    Editor = {Franklin, Michael and Kubiatowicz, John D.},
    Title = {High Availability on a Distributed Real Time Processing System},
    School = {EECS Department, University of California, Berkeley},
    Year = {2015},
    Month = {May},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-134.html},
    Number = {UCB/EECS-2015-134},
    Abstract = {Our Capstone project involves working with an open source distributed real­ time processing system called Apache Storm, in collaboration with Cisco Systems, Inc. The term “real­ time processing” in this context means that the system is able to respond within seconds or sub-­second to requests, while “distributed” means that it is running on multiple computers. The goal of the project is to add a feature called “k-­safety” to Storm. With k-­safety, Storm will be able to tolerate up to k machine failures without losing data or reducing its response time, making the system highly available. Cisco plans to integrate our modified version of Storm into their data processing pipeline and use it to support internal and customer-­facing products.}
}

EndNote citation:

%0 Thesis
%A Tanuwidjaja, Enrico
%E Franklin, Michael
%E Kubiatowicz, John D.
%T High Availability on a Distributed Real Time Processing System
%I EECS Department, University of California, Berkeley
%D 2015
%8 May 15
%@ UCB/EECS-2015-134
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-134.html
%F Tanuwidjaja:EECS-2015-134