Re: [vnfpool] Virtualized vs. physical network functions

[PEDRO ANDRES ARANDA GUTIERREZ <[e-mail]>]

Hi

I¹d really wonder why we have the Œv¹ then but let¹s assume we drop it
(could it be that the Œv¹ is the cool letter of the day?) :-) Anyhow, just
going on with what I¹m learning with the CDN use caseŠ

Let¹s assume three major categories where we want to achieve resilience:
compute, storage and network.
Leaving  the virtualised vs. physical axis of the problem aside for one
moment, let¹s look at the possible resilience mechanisms.

In the network, you can use state of the art (HSRP, etc.) to make sure two
endpoints coordinate to take care that a specific connection is not lost.
Going one step beyond that and looking at what we could do, then we could
even define a way to orchestrate the networking part and redirect traffic
towards a ¹stand-by¹ network function if there is evidence that an
Œactive¹ network function is unreachable.

As far as the compute part is concerned, we could use the same
orchestration mechanism to redirect traffic when the network function is
not responsive.

And regarding storage, we can look at two possibilities: local and remote
storage. If local storage fails, we can either treat the function as if it
was down and go back to the beginning of this paragraph. Alternatively, we
could have the NF signal a storage failure and the orchestration find an
alternative remote storage location which the NF could use to resume
working. And this same mechanism could be used to treat the case when a NF
is using remote storage only. It starts with a specific copy of the
storage and switches over to the backup copy.

In all cases, we need mechanisms to replicate the state to make sure that
the stand-by NF or storage can take over when a problem arises.

Now, going back to the Œv¹-word, and after this short recap of the
problem, I have to confess that the main difference (or advantage) of a
VNF versus a non-V NF is that you could create an instance of the NF
on-the-fly in case of failure, provided you are able to solve the state
replication issue.



Dr. Pedro A. Aranda Gutiérrez

Technology Exploration -
Network Innovation & Virtualisation

mailto:[e-mail]
Telefónica, Investigación y Desarrollo
C/ D. Ramón de la Cruz,84
28006 Madrid, Spain

Fragen sind nicht da, um beantwortet zu werden.
Fragen sind da, um gestellt zu werden.
Georg Kreisler





El 08/05/14 21:59, "Melinda Shore" <[e-mail]> escribió:

>One of the things that came up at the BOF in London and that
>hasn't received any subsequent discussion is the question of
>the applicability of network function reliability based on
>a pooling/redundancy model being applicable to network functions
>running directly on hardware, in addition to virtualized
>functions (as we've been discussing).  To be honest I'm still
>not 100% clear on the issue and it would be helpful if
>someone who's got opinions on the topic could take a whack at
>it.
>
>Thanks,
>
>Melinda
>
>--
>Melinda Shore
>No Mountain Software
>[e-mail]
>
>"Software longa, hardware brevis."
>
>_______________________________________________
>vnfpool mailing list
>[e-mail]
>https://www.ietf.org/mailman/listinfo/vnfpool


________________________________

Este mensaje se dirige exclusivamente a su destinatario. Puede consultar nuestra política de envío y recepción de correo electrónico en el enlace situado más abajo.
This message is intended exclusively for its addressee. We only send and receive email on the basis of the terms set out at:
http://www.tid.es/ES/PAGINAS/disclaimer.aspx