Re: [sami] Bringing new work into the IETF

[Thomas Narten <[e-mail]>]

"Yingjie Gu(yingjie)" <[e-mail]> writes:


> The state could be TCP states on Firewall or Session states on Load
> Balancer. When migrating state, one may find that he has to migrate
> state between devices designed by different vendors. That is why we
> need a standardized way to do this.

Well, there are 2 separate problems above. Each involves different
state, and possibly different challenges. It would be useful to ask
specifically for these two scenaros, whether there is a compelling
need to migrate such state. I.e., who needs a solution to this
problem? Which data center operator? Which  customer? Or is this just
a theoretical problem for which if there was a solution, it *might*
get used?


> Currently we are try to narrow down the scope: e.g. do we consider
> state migration within a L2 subnet, or also consider state migration
> between L2 subnets. We will also figure our the state that is
> essential to service continuity after VM Migration.

IMO, this is the less interesting question regarding scope.

The first question is whether there is a compelling case to move state
at all. If so, for which devices? And what do the vendors for those
devices say? If the vendors that own the market for firewalls,
load-balancer, <you name it but they have network state> are not
interested in a solution as is being discussed here, what is the
likelyhood that any solution (if it was developed) would ever be used?

IMO, we should start by first talking about specific scenarios, where
migrating state is necessary, and why the lack of a solution today is
a problem that the IETF needs to develop solutions for.

Thomas