CF Requirements: 4) Interoperable Suites

[Harry Mussman]

GENI Control Framework WG Members, interested GENI participants, and
Larry Lannom of CNRI,


A first DRAFT (v01.3) of the GENI Control Framework Requirements
document has been completed, and can be found at
http://groups.geni.net/geni/wiki/GeniControlFrameworkRequirements 

A discussion of the document on February 25, 2009 resulted in the
identification of 10 discussion items, which can be found at
http://groups.geni.net/geni/attachment/wiki/GeniControlFrameworkRequir
ements/030609_CFRequireReviewTopics.pdf



This email begins a discussion thread on:  4)  Interoperable Suites

I would like to invite Larry Lannom of CNRI to begin the discussion,
and then others are invited to join in.  



2a)  During the discussion, the number and nature of interoperable
GENI suites was discussed.

In his earlier comments on the document, Larry said that he wasn't
sure if there was one GENI suite, or multiple interoperable suites.
And was GENI also the name of an interoperability spec?  


2b)  The current DRAFT in Sec 5.7.2 says:  
" The control framework shall provide for the federation of a GENI
suite with one or more suites that utilize the same control framework
structure as the GENI suite.
For example, the federation of an NSF-sponsored GENI suite with an
EU-sponsored GENI suite.
For example, the federation of an NSF-sponsored GENI suite with twenty
university-sponsored GENI suites."

However, the system overview looks like one big NSF-sponsored GENI
suite, plus a federated EU-sponsored GENI suite.
There is very little said about the interoperability interfaces that
would be required for multiple GENI suites.


2c)  Discussion:
Considering the federation approach that has been done in PlanetLab
and Emulab to date, it is reasonable to imagine that there will be
multiple GENI suites at the start, associated with research
organizations, i.e., universities.
Certainly we can expect multiple suites split across diverse regions,
i.e., US, Europe and Asia.
Also, may well have multiple suites in the US, where some smaller
suites are dedicated to a research organization (or group), and these
are all federated with each other via an "NSF suite".  
Expect much of the traffic to stay within a smaller suite, i.e., most
of the traffic associated with a teaching class.
Considering the face that the GENI Spiral 1 has 5 control frameworks,
you could say that we are starting with 5 GENI suites.
Even if there was only one suite, we would need multiple entities to
reflect redundancy, and permit updates without flash cuts.


2d)  A proposed change to the CF requirements:
This document should clearly reflect that there are several GENI
suites.
Interfaces that allow them to interoperate must be fully defined.  
There is not "one big suite".  
This should be reflected in the System Overview, and not just
mentioned in the later Federation section.


2e)  Continued discussion:
How can we define a precise interoperability spec between GENI suites?
Is there room for more than one spec at the start?
Between which elements?
Between "Slice Controller" and "Aggregate Manager"?
Between "Clearinghouses"?
What are the minimal interfaces required to permit interoperability
between GENI suites?  
What are the minimal functions required to permit interoperability
between GENI suites? (there will be a separate discussion thread on
this topic)
What is the nature of the spec between "Slice Controller" and
"Aggregate Manager"?
Based on certificates, like PlanetLab and ProtoGENI?
Based on web services, like ORCA?  
Fully defined, so easy to realize interface code?
What is the nature of the spec between "clearinghouses"?  
Like web services?  
Fully defined, so easy to realize interface code?


We look forward to a continuing, lively discussion, working towards a
'rough consensus".

Harry E. Mussman
Control Framework Systems Engineer

GENI Project Office
BBN Technologies
10 Moulton Street
Cambridge, MA 02138
[phone] - Office
[phone] - Mobile
[phone] - Fax
hmussman@...
www.bbn.com



_______________________________________________
control-wg mailing list
control-wg@...
http://lists.geni.net/mailman/listinfo/control-wg