Add package NSID3V (scalable NS)

[sonata-nfv]

Add package NSID3V (scalable NS)

[felipevicens]

Do we need a special VM image?

[mpeuster]

I do not know. Basically I do not know where to start at all. Never used scaling with the SP. I do not even know how a descriptor for this has too look like ... I think someone from WP5 needs to help here.

[carlos-f-parada]

Hi @mpeuster,
Don't know the full context, but automatic (Policy-based) or Manual (API-based) scaling are supported. I left you some pointers.
For manual scaling you can check the API here https://github.com/sonata-nfv/tng-gtk-sp/wiki/Service-scaling-out-or-in.
For automatic, you need to create a Policy to that service and to apply it (@efotopoulou can give you more details).
In both cases you need an SSM to catch the scaling request (@tsoenen can help you on that).

[mpeuster]

The question I would have is how should the NSD and VNFD for this scenario should look like? So we could provide such an example package? Is it a normal service, or do we need anything special? Lets assume the manual scaling case for now.

[carlos-f-parada]

Scaling (manual or automated) there are no impacts on Descriptors, they are scaling unaware. Scaling is tacked outside the descriptors scope.

[jbonnet]

But @carlos-f-parada can a NS scale out infinetly? I think the impact on the descriptors would be somehow related to this, the limits to which it can scale out/in... ARe those limits covered?

[carlos-f-parada]

@jbonnet I think you can define limits for scaling, but not sure if that is in use. No sure if the MANO consider these values when the Policy Manager requests a scaling action to the MANO. Only @tsoenen can tell us.

[mpeuster]

Ok, but why need we a dedicated test package for this then @felipevicens ? Why not just use the existing ones?

[felipevicens]

Hey Manuel, good question. I think because the image of the VNF to apply a policy for scaling. I'll define the scenario and if we need something special I let you know, if not I'll close this issue :)

[mpeuster]

Perfect 👍

[tsoenen]

A policy is based on a monitoring metric. We can use the same VM images, as long as we define a metric that we can push above a threshold.