HP 200 Series Services And Applications page 188

Data Compression for WAN Links
Compression Performance Testing
Compression Performance Testing
HP conducted performance testing using the Calgary Corpus test files,
which are industry-standard files for performance testing using various
types of data. The table below describes each file in the Calgary Corpus set.
File name
bib
book1
book2
geo
news
obj1
obj2
paper1
paper2
paper3
paper4
paper5
paper6
*pic
progc
progl
progp
trans
* The pic file was left out of the testing because it contains mostly zeros, which skewed
the compression results in favor of higher rations. For example, just copying the pic file
by itself yielded approximately 5:1 compression ratio. By leaving out the pic file, we
achieved a file set that more closely approximated test files.
2-136
File Description
Bibliographic files
Hardy: Far form the Madding Crowd
Witten: Principles of Computer Speech
Geophysical data
News batch file
Compiled code for VAX: compilation of progp
Compiled code for Apple Macintosh: knowledge
support system
Witten, Neal and Cleary: Arithmetic coding for
data compression
Witten: Computer (in) security
Witten: In Search of "Autonomy"
Cleary: Programming by Example Revised
Cleary: A Logical Implementation of Arithmetic
Cleary: Compact Hash Tables Using Bidirectional
Linear Probing
Picture number 5 from the CCITT facsimile test
files (test + srawings)
C source code: compress version 4.0
LISP source code: system software
Pascal source code: prediction by partial
matching evaluation program
Transcript of a session on a terminal
File Size (bytes)
111261
768771
610856
102400
377109
21504
246814
53161
82199
46526
13286
11954
38105
513216
39611
71646
49379
93695