The question you wrote in the end, is exactly my question:)
Should pings/KPI/etc be conducted by ODL, or should this logic reside in an 
external monitoring tool, and notified back to the ODL on state change?

From: Brady Allen Johnson []
Sent: Tuesday, September 20, 2016 11:02 AM
To: Sela, Guy <>;; controller-dev 
Subject: Re: [controller-dev] Health Checks to external network functions

The SFC project has functionality that can "indirectly" monitor network 
functions (called Service Functions in SFC). I say indirectly since the 
monitoring is used for what we call Service Function (SF) scheduling 
algorithms, which helps us choose which SF to use based on either CPU load, or 
shortest path.

When we integrated ODL SFC into OPNFV, we didnt use the SF scheduling 
algorithms but instead delegated that to a VNF manager. Typically, the VNF 
manager would also monitor the network functions.

In Beryllium, the Armoury project was created to monitor the network load of a 
network function which could then report overloading situations to a VNF 
manager. The Armoury project didnt get much traction, and is effectively on 
hold for now.

The question needs to be asked if it really makes sense for ODL to monitor 
SFs/Network Functions or if something like a VNF manager should do it instead.



On 20/09/16 07:31, Sela, Guy wrote:

Is there any project/plugin in ODL that its purpose is to conduct health checks 
I'm referring to the health of external machines that run network functions for 
example (firewalls, load balancers, content filtering, etc.).
Health checks can be pings, load, generic scripts etc.

In a real deployment, will you do those inside ODL? or use an external health 
tool that will notify status change to the ODL?

Guy Sela


controller-dev mailing list<>

controller-dev mailing list

Reply via email to