Hi,
It’s a little bit complicated topic. In fact, in the bar metal approach, a single vendor is managing the whole E2E stack; from Hardware to middleware to application to the final service. Today, within the virtualization, the NFVi layer should provide specific SLA for each application (VNF).
I use to include in my VNF RFP some sort of table to help me calculated the RTO and RPO per VNF component.
In fact, I use to do that within the IT applications and it’s a best practice that I have inherit from ISO 22301 for business continuity.
RPO: Recovery time objective
RTO: Recovery point objective
These two factors are calculated per time unit. ( minute)
So, as a first step, I try to analyze the dependency between the VNF components:
|
|
VNF-Component
|
VNF-C1
|
VNF-C2
|
VNF-C3
|
VNF-C4
|
VNF-Cn
|
Functional Impact
|
|
VNF 1
|
VNF-C1
|
|
|
|
|
|
|
|
VNF-C2
|
|
|
|
|
|
|
|
VNF-C3
|
|
|
|
|
|
|
|
VNF-C4
|
|
|
|
|
|
|
|
VNF-Cn
|
|
|
|
|
|
|
|
|
VNF-Component
|
VNF-C1
|
VNF-C2
|
VNF-C3
|
VNF-C4
|
VNF-Cn
|
Functional Impact
|
|
VNF 2
|
VNF-C1
|
|
|
|
|
|
|
|
VNF-C2
|
|
|
|
|
|
|
|
VNF-C3
|
|
|
|
|
|
|
|
VNF-C4
|
|
|
|
|
|
|
|
VNF-Cn
|
|
|
|
|
|
|
The idea behind that analyze is to find the most critical component in a specific VNF.
Today, if we take the example of a vIMS: it can be composed from different VNFs: CSCF, TAS…
Inside each VNF, you have different component, the operator should be aware of the criticity of each component and how this component impact the whole service availability.
Using that, you will be able to define the dependency between all of these components in term of service availability and resiliency.
Then, I ask application vendors to specify the impact in term of time and data of the no availability of each component.
|
|
VNF-Component
|
Resiliency Scheme
|
RPO (min)
|
RTO (min)
|
Restart Sequence
|
|
VNF 1
|
VNF-C1
|
|
|
|
|
|
VNF-C2
|
|
|
|
|
|
VNF-C3
|
|
|
|
|
|
VNF-C4
|
|
|
|
|
|
VNF-Cn
|
|
|
|
|
|
|
VNF-Component
|
Resiliency Scheme
|
RPO (min)
|
RTO (min)
|
Restart Sequence
|
|
VNF 2
|
VNF-C1
|
|
|
|
|
|
VNF-C2
|
|
|
|
|
|
VNF-C3
|
|
|
|
|
|
VNF-C4
|
|
|
|
|
|
VNF-Cn
|
|
|
|
|
RPO and RTO for each VNF Component shall be specified along with VNF-Cs restart sequence if applicable and the resiliency mechanism of the VNF-C. An example is given in the following table.
|
|
VNF-Component
|
Resiliency Scheme
|
RPO (min)
|
RTO (min)
|
Restart Sequence
|
|
VNF
|
VNF-C1
|
N+1
|
60
|
2
|
After VNF-C2
|
|
VNF-C2
|
N+1
|
60
|
2
|
After VNF-C4
|
|
VNF-C3
|
1+1
|
30
|
5
|
First
|
|
VNF-C4
|
N+M
|
15
|
10
|
After VNF-C3
|
|
VNF-Cn
|
N+1
|
30
|
2
|
After VNF-Cn-1
|
After defining all of the above elements, you are now able to map all of these information to your infrastructure that will host your VNFs.
Usually, your infrastructure should be multi-tenancy and gives the application the possibility to define different clusters (pool) of resources within different resiliency and availability matrix.
The virtualized infrastructure, will help the application to provide the needed SLA from service point of view.
Actually many Telco vendor, are relying on the application resiliency mechanism in a virtualized environment unless the VIM layer and MANO layer are handling HA from their side.
Cheers,
Houcem
------------------------------
Houcemeddine GARBOUJ
Cloud NFV/SDN Architect
gmail
nabeul
+33605604805
------------------------------
Original Message:
Sent: 01-30-2017 15:31
From: Marco Carosi
Subject: SLA management in NFV infrastructure
I'm interested in understanding how to implement the SLA management in the NFV infrastructure. How can you calculate the availability of the infrastructure toward the VNF? I’m interested in possible algorithms and tools.
------------------------------
Marco Carosi
Service Manager
Telecom Italia Group
roma
------------------------------