Check out the new USENIX Web site.

next up previous
Next: Detection Results Up: Performance Evaluation Previous: Performance Evaluation

Experimental Setup

The experiments were conducted with FreeBSD 2.2.5 on a 166MHZ Intel Pentium PC with 64MB RAM and a 2.1GB Quantum Fireball hard disk. The applications we used are described below and are summarized in Table 1.

Table 1: Characteristics of the applications.
Application Description Input data (MB)
cscope C examination tool C code (9)
glimpse information retrieval tool text files (50)
sort UNIX sort utility text files (4.5)
link UNIX link editor object files (2.5)
cpp C preprocessor C code (11)
gnuplot GNU plotting utility numeric data (8)
postgres1 relational DB system two relations
postgres2 relational DB system four relations
Cscope is an interactive C-source examination tool. It creates an index file named cscope.out from C sources and answers interactive queries like searching C symbols or finding specific functions or identifiers. We used cscope on kernel sources of roughly 9MB in size and executed queries that search for five literals.
Glimpse is a text information retrieval utility. It builds indexes for words and allows fast searching. Text files of roughly 50MB in size were indexed resulting in about 5MB of indexes. A search was done for lines that contain the keywords multithread, realtime, DSM, continuous media, and diskspace.
Sort is a utility that sorts lines of text files. A 4.5MB text file was used as input, and this file was sorted numerically using the first field as the key.
Link is the UNIX link-editor. We used this application to build the FreeBSD kernel from about 2.5MB of object files.
Cpp is the GNU C-compatible compiler preprocessor. The kernel source was used as input with the size of header files and C-source files of about 1MB and 10MB, respectively.
Gnuplot is a command-line driven interactive plotting program. Using 8MB raw data, we plotted three-dimensional plots four times with different points of view.
postgres1 and postgres2
Postgres is a relational database system from the University of California at Berkeley. PostgresSQL version 6.2 and relations from a scaled-up Wisconsin benchmark such as thoustup and tenthoustup were used. Postgres1 is a join between the hundredthoustup and twohundredthoustup relations while postgres2 is a join among four relations, namely, fivehundredup, twothoustup, twentythoustup, and twohundredthoustup. The sizes of fivehundredup, twothoustup, twentythoustup, hundredthoustup, and twohundredthoustup are approximately 50KB, 150KB, 1.5MB, 7.5MB, and 15MB, respectively.

next up previous
Next: Detection Results Up: Performance Evaluation Previous: Performance Evaluation

Jongmoo Choi