Re: [tosca] Re: Differentiating TOSCA from HEAT

[Tal Liron <tliron@redhat.com>]

On Sun, Jan 30, 2022 at 4:59 PM Chris Lauwers <lauwers@ubicity.com> wrote:

In my templates, I provide exactly this functionality, but rather than using âtemplatingâ in my templates, I use substitution mapping instead. This aligns philosophically better with âintent-based orchestrationâ anyway. At the highest level of abstraction, a user-provided input will specify if âload balancingâ is required or not (similar to Peterâs earlier example where someone would specify if âfirewallingâ is required or not). The orchestrator will then substitute either a template that includes a load balancer, or a template that includes a web frontend.

Â

By the way, this is exactly what ETSI NFV does as well: they provide variability in VNFs by using substitution mapping to deploy different deployment flavors.

This is the right fundamental approach to variability: keep it strictly designed and tested with actual topologies, which are themselves complete.

But I do think we can make TOSCA better in this regard. Currently substitution is limited to a one-template mechanism. This is good for ETSI NFV because they have a rigid hierarchy of network services comprising network functions. But not everybody, even in the NFV world, follows ETSI's idealized architecture. I think this approach is just not good enough for many reasonable use cases, even in telco.

I will continue to encourage us to think about service composition more broadly.