Subject: Sensor server proxy to Ganglia
Reply-To: "Brent N. Chun" <bn...@in...>
Organization: Intel Research Berkeley
--text follows this line--
I've just hacked up and deployed a sensor server proxy to Ganglia on
PlanetLab. The server listens on TCP port 2841 and returns tuples as
shown below:
planet1% curl -s http://127.0.0.1:2841/ganglia/README
This sensor returns a list of tuples, each of which corresponds to
a set of metrics for a particular host at a particular PlanetLab
site. At most PlanetLab sites, there are typically 2-3 nodes and
hence there will be 2-3 tuples returned. Each tuple has the
following fields:
site PlanetLab site name
ip Node IP address
hostname Node hostname
boottime
bytes_in
bytes_out
cpu_aidle
cpu_idle
cpu_nice
cpu_num
cpu_speed
cpu_system
cpu_user
disk_free
disk_total
gexec
load_fifteen
load_five
load_one
machine_type
mem_buffers
mem_cached
mem_free
mem_shared
mem_total
mtu
os_name
os_release
part_max_used
pkts_in
pkts_out
proc_run
proc_total
swap_free
swap_total
sys_clock
Here's some sample output:
planet1% curl http://127.0.0.1:2841/ganglia
Intel Research Berkeley, 12.46.129.21, planet1.berkeley.intel-research.net, 1060992195, 87868.00, 206.73, 75.2, 88.6, 0.0, 1, 1263, 3.8, 23.5, 22.375, 67.968, OFF, 0.46, 0.40, 0.63, x86, 113152, 402804, 13884, 0, 905012, 1500, Linux, 2.4.19-6_planetlab, 100.0, 32315682.00, 236.52, 0, 157, 1915708, 2040244, 1061865729
Intel Research Berkeley, 12.46.129.22, planet2.berkeley.intel-research.net, 1060992652, 25539.19, 114.42, 73.8, 0.0, 0.7, 1, 1263, 17.7, 82.1, 23.736, 67.968, OFF, 3.01, 2.95, 2.72, x86, 200536, 195464, 5220, 0, 905012, 1500, Linux, 2.4.19-6_planetlab, 100.0, 106231984.00, 114.35, 3, 143, 1965008, 2040244, 1061864886
Intel Research Berkeley, 12.46.129.23, planet3.berkeley.intel-research.net, 1060988740, 2868.60, 29.76, 87.8, 98.0, 0.0, 1, 1263, 0.3, 2.2, 40.909, 67.936, OFF, 0.23, 0.18, 0.18, x86, 62148, 228388, 537920, 0, 905012, 1500, Linux, 2.4.19-6_planetlab, 100.0, 29.71, 95443656.00, 0, 115, 2015348, 2040244, 1060988871
The set of nodes where the server is currently running can be found
here. It currently runs on 144 PlanetLab nodes:
http://berkeley.intel-research.net/bnc/ganglia/nodes.txt
The proxy, combined with Sophia/PIER/etc. should finally allow us to
use the data Ganglia collects for performing some interesting resource
discovery queries.
bnc
|
