[Planetlab-devel] Re: [Fwd: Re: cache-based federation and slice attributes]

[Stephen Soltesz]

Hey, guys,

I hope it's ok that this goes back to the devel list.  It looks like it fell 
off accidentally.

The 'service creation' idea is very similar to the conversation at the PL 
meeting a few weeks ago.  You might remember it as Plan A and Plan B, or 
Larry and Marc discussing approaches to federation.

The exchange of slices is the primary mechanism of federation. So, for 
comparison, imagine the current federation approach where the SA of PLC1 
sends all of its slices to the MA of PLC2.  The existing federation 
agreement implies that PLC2 will instantiate all of these slices on behalf 
of PLC1.  One concern is that this collects all of the slice information in 
the MA.  This could have scalability concerns.

The 'service creation' idea is just a generalization of this 'send all 
slices' approach, where instead of all slices at one extreme, the opposing 
extreme would instead send only a single slice.  Of course, a single slice 
is not very exciting from a federation perspective.  So, the idea is that 
PLC1's SA can create additional slices by running a simple Slice Creation 
Service (SCS) that would talk to local NM on PLC2's machines.  The resources 
for instantiating additional slices that the SCS created would be 'carved 
off' of the single 'service creation' slice.

The ultimate idea is to allow for intermediate combinations of the two.  For 
example a single MA peers with two different SAs, and SA1 gets a dozen 
slices, but SA2 only gets a single slice.  This now becomes a policy 
decision to be decided during the creation of the peering agreement between 
the SA and MA of the different sites.  The MA is protected from either SA 
going crazy with its database.  And I think it is appropriate to think of 
the situation as pushing some of the complexity from the MA to the nodes.

The nodes in this scenario remain a cache for policy maintained by either 
the SA or MA.

Does this help clarify?  This may not be everything, so please let me know 
if some part is not clear.
Stephen.

Thierry Parmentelat wrote:
> Stephen, do you want to shoot first, or do you prefer me to ?
> 
> 
> -------- Original Message --------
> Subject:     Re: cache-based federation and slice attributes
> Date:     Tue, 17 Jul 2007 11:53:19 -0400
> From:     Faiyaz Ahmed <faiyaza at CS.Princeton.EDU>
> To:     Thierry Parmentelat <Thierry.Parmentelat at sophia.inria.fr>
> References:     <469CDFE8.9060500 at sophia.inria.fr>
> 
> 
> 
> Hi Theirry,
> 
>> As a final note. The idea that was presented by Stephen in Warsaw, 
>> about using a 'delegation' slice on foreign nodes as a 'service 
>> creation handle' looks real nice, and the more I think about it the 
>> more I like this idea; it would clearly have a simpler impact on the 
>> data model than the current cache-based federation scheme had had, and 
>> would make the separation between MA and SA much more natural. It 
>> would be however, mostly orthogonal with the issue I described above; 
>> IMHO.
> 
> For the benefit of those who weren't at the Warsaw meeting, can you (or 
> Stephen) briefly describe the "service creation" idea.  I'd like to make 
> sure we're all on the same page.
> 
> Thanks.
> 
> 
> Faiyaz