- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Implementation of processor farms on myrinet network...
Open Collections
UBC Theses and Dissertations
UBC Theses and Dissertations
Implementation of processor farms on myrinet network interface cards Parsa, Marjan
Abstract
Processor farms are an intuitive way of parallelizing a program in which a manager processor farms out tasks to a set of worker processors that process the tasks and send the results back to the manager. In this thesis we consider processor farms as a possible intermediate level model which is sufficiently high level to be easy to use, but simple enough to run efficiently on the latest cluster architecture. Advances in programmable interconnects have made it possible to increase the efficiency of processor farms. These interconnections have a programmable processor which enables us to offload the communication to the boards. Also, due to the recent rapid increase in processor speeds, using high speed interconnections allows the communication to keep up with the computation. The experimental test bed for our processor farm system is a cluster of sixteen Pentium II computers interconnected by a Myrinet [18] gigabit per second programmable network. We developed and optimized a processor farm, which we call pfarm for our sixteen node cluster. We show that pfarm effectively executes relatively fine grain tasks with variable execution times, using a programmable network interface card to almost completely overlap communication and computation.
Item Metadata
Title |
Implementation of processor farms on myrinet network interface cards
|
Creator | |
Publisher |
University of British Columbia
|
Date Issued |
1998
|
Description |
Processor farms are an intuitive way of parallelizing a program in which a manager
processor farms out tasks to a set of worker processors that process the tasks and
send the results back to the manager. In this thesis we consider processor farms as
a possible intermediate level model which is sufficiently high level to be easy to use,
but simple enough to run efficiently on the latest cluster architecture.
Advances in programmable interconnects have made it possible to increase
the efficiency of processor farms. These interconnections have a programmable processor
which enables us to offload the communication to the boards. Also, due to the
recent rapid increase in processor speeds, using high speed interconnections allows
the communication to keep up with the computation. The experimental test bed for
our processor farm system is a cluster of sixteen Pentium II computers interconnected
by a Myrinet [18] gigabit per second programmable network. We developed
and optimized a processor farm, which we call pfarm for our sixteen node cluster.
We show that pfarm effectively executes relatively fine grain tasks with variable
execution times, using a programmable network interface card to almost completely
overlap communication and computation.
|
Extent |
2904548 bytes
|
Genre | |
Type | |
File Format |
application/pdf
|
Language |
eng
|
Date Available |
2009-05-26
|
Provider |
Vancouver : University of British Columbia Library
|
Rights |
For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.
|
DOI |
10.14288/1.0051378
|
URI | |
Degree | |
Program | |
Affiliation | |
Degree Grantor |
University of British Columbia
|
Graduation Date |
1998-11
|
Campus | |
Scholarly Level |
Graduate
|
Aggregated Source Repository |
DSpace
|
Item Media
Item Citations and Data
Rights
For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.