Demonstrations of cachestat, the Linux ftrace version.


Here is some sample output showing file system cache statistics, followed by
the workload that caused it:

# ./cachestat -t
Counting cache functions... Output every 1 seconds.
TIME         HITS   MISSES  DIRTIES    RATIO   BUFFERS_MB   CACHE_MB
08:28:57      415        0        0   100.0%            1        191
08:28:58      411        0        0   100.0%            1        191
08:28:59      362       97        0    78.9%            0          8
08:29:00      411        0        0   100.0%            0          9
08:29:01      775    20489        0     3.6%            0         89
08:29:02      411        0        0   100.0%            0         89
08:29:03     6069        0        0   100.0%            0         89
08:29:04    15249        0        0   100.0%            0         89
08:29:05      411        0        0   100.0%            0         89
08:29:06      411        0        0   100.0%            0         89
08:29:07      411        0        3   100.0%            0         89
[...]

I used the -t option to include the TIME column, to make describing the output
easier.

The workload was:

# echo 1 > /proc/sys/vm/drop_caches; sleep 2; cksum 80m; sleep 2; cksum 80m

At 8:28:58, the page cache was dropped by the first command, which can be seen
by the drop in size for "CACHE_MB" (page cache size) from 191 Mbytes to 8.
After a 2 second sleep, a cksum command was issued at 8:29:01, for an 80 Mbyte
file (called "80m"), which caused a total of ~20,400 misses ("MISSES" column),
and the page cache size to grow by 80 Mbytes. The hit ratio during this dropped
to 3.6%. Finally, after another 2 second sleep, at 8:29:03 the cksum command
was run a second time, this time hitting entirely from cache.

Instrumenting all file system cache accesses does cost some overhead, and this
tool might slow your target system by 2% or so. Test before use if this is a
concern.

This tool also uses dynamic tracing, and is tied to Linux kernel implementation
details. If it doesn't work for you, it probably needs fixing.


Use -h to print the USAGE message:

# ./cachestat -h
USAGE: cachestat [-Dht] [interval]
                 -D              # print debug counters
                 -h              # this usage message
                 -t              # include timestamp
                 interval        # output interval in secs (default 1)
  eg,
       cachestat                 # show stats every second
       cachestat 5               # show stats every 5 seconds

See the man page and example file for more info.