home *** CD-ROM | disk | FTP | other *** search
- The data files contained in this directory accompany the showcase
- files up one directory in toolbox/perfTuning/ChallengeIO.
-
- The files are:
-
- a) lvres.* files are the logs of multiple runs of sio.
- These files show the sio parameters used for the runs.
-
- 1) 1vs4 shows 1 segment versus 4 segments.
- 2) bvst shows write buffered versus write through.
- 3) step shows varying step size performance.
-
- b) graph_* are scripts that process lvres.* files and average
- the numbers from the same runs and produce tables
- of the performance (the perf.* files) and shell scripts
- that invoke xgraph to display the data graphically
- (graph.*).
- The graph_* files generate the perf.* and graph.* files
- from the corresponding lvres.* files.
-
- c) perf.* files contain tables showing varying read/write/create
- performance with raw/direct/efs.
-
- - From graph_1vs4 lvres.1vs4 - 1 versus 4 segment
- 1) dw1 - direct, 1 segment
- 2) dw4 - direct, 4 segment
- 3) ew1 - efs, 1 segment
- 4) ew4 - efs, 4 segment
- 5) rw1 - raw, 1 segment
- 6) rw4 - raw, 4 segment
-
- - From graph_bvst lvres.bvst - write buffered vs write through
- 1) dwb - direct, write buffered
- 2) dwt - direct, write through
- 3) ewb - efs, write buffered
- 4) ewt - efs, write through
- 5) rwb - raw, write buffered
- 6) rwt - raw, write through
-
- - From graph_step lvres.step - step size variation (1 vs 4 seg)
- 1) 1seg - 1 segment
- 2) 4seg - 4 segment
-
- d) graph.* files are shell scripts that invoke xgraph with the
- appropriate data, titles, axes, etc. The following
- are present:
-
- - From graph_1vs4 lvres.1vs4 - 1 versus 4 segment
- 1) dw1 - direct, 1 segment
- 2) dw4 - direct, 4 segment
- 3) ew1 - efs, 1 segment
- 4) ew4 - efs, 4 segment
- 5) rw1 - raw, 1 segment
- 6) rw4 - raw, 4 segment
- 7) c - create performance for direct and efs.
- 8) r - read performance for raw, direct and efs.
- 9) w - write performance for raw, direct and efs.
-
- - From graph_bvst lvres.bvst - write buffered vs write through
- 1) dwb - direct, write buffered
- 2) dwt - direct, write through
- 3) ewb - efs, write buffered
- 4) ewt - efs, write through
- 5) rwb - raw, write buffered
- 6) rwt - raw, write through
- 7) bc - create performance for direct and efs.
- 8) br - read performance for raw, direct and efs.
- 9) bw - write performance for raw, direct and efs.
-
- - From graph_step lvres.step - step size variation (1 vs 4 seg)
- 1) sc - create performance for 1 and 4 segment.
- 2) sr - read performance for 1 and 4 segment.
- 3) sw - write performance for 1 and 4 segment.
-
- e) ps.* files are postscript files for the graphs drawn by the
- graph files.
-
- 1) The ps.14.* files are the graphs generated from
- "graph_1vs4 lvres.1vs4".
- 2) The ps.bt.* files are from "graph_bvst lvres.bvst".
- 3) The ps.24.* files are from "graph_step lvres.step",
- and are the results of varying step sizes on a
- 24-way stripe with 1 and 4 cache segments.
-
-