Slice monitoring sensor on PlanetLab
Slicestat provides slice-level
resource consumption information on each PlanetLab node. It is a
tiny sensor which takes the snapshot of current resource usage of
active slices on a node. The original version was developed by Brent Chun, but later
we rewrote it for PlanetLab version 3. In transitioning to the
current version, we made the server event-driven, and added support
for request coalescing. This merges the simultaneous requests to
wait for one pending query, thus lessening the burden of excessive
Slicestat exports the following information, separated by commas, in one line per slice. The CPU and memory usage information is earned from /proc(e.g., 'top') while the bandwidth information is gleaned from the Scout sensor.
ex: ... mit_dht,636,0.0,0.5,5416,21980,3,0.000000,0.026605,0.026605, 0.000000,0.226073,0.226073,220.127.116.11,0 ...
Slicestat communicates in HTTP format
on port 3100. You can use any HTTP client tools(e.g., curl, wget
etc.). As in the original version, it supports the "max-age" header
to request a cached result. For example, "max-age: 5" means the
result is acceptable if it is less than five seconds old.
ex: curl -s -H "max-age: 5" http://localhost:3100/slicestat (on any PlanetLab node)
curl -s http://planetlab-1.cs.princeton.edu:3100/slicestat
Please refer to the CoMon status page.
the nice web interface for slice information as well as
node-centric view of the resource consumption status.
Here is the source code(Please take
a look at costat.c which contains the main logic). In order to
compile, you need to install the Proper
Please send your feedback and any bug report to us. We may collectively be contacted at princeton_slicestat at slices.planet-lab.org