Name | Organization | |
Jian Li | POSTECH ON.Lab |
Control Plane Manager (CPMan) adds management capability to control plane in a way to provide higher availability and reliability for ONOS.
The features of CPMan are comprised of two parts: 1) control plane monitoring, 2) enhanced mastership management.
This subsystem is used to aggregate and collect various types of control message from network devices. Followings are the components of the subsystem.
Following table shows the mapping between abstracted message type and OpenFlow message type.
Abstracted Message Type | OpenFlow Message Type |
---|---|
INBOUND_PACKET | PACKET_IN |
OUTBOUND_PACKET | PACKET_OUT |
FLOW_MOD_PACKET | FLOW_MOD |
FLOW_REMOVED_PACKET | FLOW_REMOVED |
REQUEST_PACKET | STATS_REQUEST |
REPLY_PACKET | STATS_REPLY |
Before you start, you will need followings:
Install and run ONOS instance. The detailed procedures can be refer from here.
To collect control message, we also need to deploy a set of devices and connect to ONOS.
This can be realizable by setting up an OpenFlow network using mininet.
Mininet installation and configuration can be referred from http://mininet.org/download/.
After ONOS and mininet get initialized, we can activate CPMan application.
onos > app activate org.onosproject.cpman |
As long as CPMan is activated, it automatically senses control messages and stores stats. into local RRD.
If you want to collect system metrics (e.g., CPU, memory usages, disk I/O, etc.), please refer to this link.
The statistics information of control metrics can be queried through CLI.
In ONOS console, you can type following command to view control message stats.
onos > cpman-stats-list <node_ip> control_message <device_id> |
You also can query other control metrics by specifying the control metric type such as cpu, memory, disk, network, etc.
CPMan also provides a way to visualize the control metrics using bar/line chart.
You can find the CPMan GUI under network category in ONOS GUI.
By default, CPMan shows a bar chart of control message stats. per device.
A detailed time-series line chart can be visualizable by clicking each bar chart.
If you already have your own monitoring system, you may want to export ONOS control metrics to third-party system.
ONOS provides two ways to support metrics export.
The first way is pull based approach with which you need to periodically pull the control metrics from your system through ONOS REST API.
The detailed REST API can be referred from this link.
The second way is push based approach with which ONOS periodically report a set of control metrics to predefined third-party system.
By far, ONOS supports three systems and those are InfluxDB, ganglia and graphite monitoring system.