CLUE: A Cluster Evaluation Tool Page: 25
View a full description of this thesis.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
information from node 3. Each node requires only the values contained in the adjacent edge
of their neighbors. Node 1 only utilizes the information on the row labeled B3 from node 2,
and likewise node 2 requires only the row labeled B2 from node 1. To expedite this sharing
of values and optimize network buffering, the CLUE benchmark cellular automata algorithm
has each node trade the entire edge row with the corresponding neighboring nodes at the
start of each grid transition iteration. Each node then stores its neighbors' edge rows in a
local memory cache to avoid redundant value requests from other nodes. This methodology
has been found to increase performance in parallel environments [16],[4].
For demonstration purposes, the grid in Figure 3.1 was split as shown in Figure 3.2
keeping the same overall 16x16 size. This methodology would follow the speedup model in
accordance with Amdahl's law. The CLUE benchmark, however, increases the overall data
size linearly with the increase in processing nodes, such that in the example of Figure 3.1,
instead of splitting the grid in fourths, each of the nodes would obtain an entire 16x16 grid.
This follows the speedup model described by Gustafson, which is model used for calculations
throughout the CLUE benchmark [7].
3.2. Designing a Good Benchmark
The premise for CLUE is straightforward: utilize a simple algorithm to test system scal-
ability. Then, alter aspects of the application and compare the results of the altered version.
The difference in performance due to the alteration then helps identify how certain system
components affect overall performance. Figure 3.3 diagrams the application alterations and
comparisons that can be made.
The CLUE benchmark consists of two applications: a C++ cellular automata timing test,
and a Perl script. The user executes the Perl script which analyzes the system to determine
ideal parameters, then begins to execute the C++ timing tests with different parameters. The
timing tests return comma separated lists of values for the executions. Although the values
obtained directly from the C++ program are fully usable for manual analysis, the Perl scripttakes these values and performs a comparative analysis between the timing tests and creates
25
Upcoming Pages
Here’s what’s next.
Search Inside
This thesis can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Thesis.
Parker, Brandon S. CLUE: A Cluster Evaluation Tool, thesis, December 2006; Denton, Texas. (https://digital.library.unt.edu/ark:/67531/metadc5444/m1/33/: accessed March 28, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; .