CF Requirements: 6) Transactions

[Harry Mussman]

GENI Control Framework WG Members, and Dave Irwin of UMass Amherst,


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:  5)  Transactions

I would like to invite Dave Irwin of UMass Amherst to begin the
discussion, and then others are invited to join in.  



6a)  On the call, Harry indicated that the basic transaction that the
CF needs to support is that researchers can obtain resources, and then
control them to execute experiments.  

Dave indicated that ORCA is based on two tenets:  resources are
shared, but not forever.

The group then had a wide-ranging discussion on the nature of
transactions, and felt that the CF needs to support a range of
transactions.


6b)  The current DRAFT of the CF Requirements document covers a
variety of situations in Section 5.5.4.

But, this section is a somewhat confusing listing of options, without
a clear structure.

 
6c)  Perhaps we should instead consider this approach:

        The CF allows the authorization and assignment of resources
from aggregates to researchers...

        through negotiation of a "lease" (or "ticket", but currently
ticket implies a certificate) that includes:

                characteristics and/or identity of resource

                time (can be now, specific, or indefinite) and
duration (must be limited)

                level of assurance

        with ability for aggregate to change negotiated "lease"

        with ability for researcher to change negotiated "lease"




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

[David Irwin-3]

Hi Harry,

FIrst off, I'm not sure if I like the word "Transaction" for this
topic, given the power and meaning often conveyed by that term.  I
think what we're talking about here are temporary resource exchanges
or transfers, negotiated between Clearinghouse (and/or Aggregate
Manager) and Slice Controller.

I generally agree with your characterization below and in the document.

In my mind, the basic sequence is as follows (which seems to also
follow the proposed GENI Architecture):   The Slice Controller issues
a request to the Clearinghouse (and/or AM) for some resources.  The
request need not be specific; i.e, it could be that I want something
sometime in the next 5 days, or I want between 1 and 100 of resource
X, etc.    The Clearinghouse responds with some allocation that meets
the criteria (or it could make a counter-proposal if resources within
the limits aren't available), which includes the set of resources and
the duration of access (lease) in a ticket.    The Slice Controller is
free to go back to the Clearinghouse at anytime to change the
parameters of the lease (resource set, duration, etc.), by going
through a similar re-negotiation/re-ticketing process.

This basic procedure does not preclude the range of transactions or
exchanges you cite in the document (pasted in below).

Resources that range from:  valuable to not valuable.
Scope of resources from:  wide (many) to narrow (few).
Resource control depth from:  superficial to deep (detailed).
Researchers that range from:  well known insiders to possibly
anonymous outsiders.
Access control that ranges from: rigourous to loose
Quality of service that ranges from: reliable to casual.

The lease is critical for using valuable resources:  it seems unlikely
someone will transfer control of a valuable resource without knowing
their commitment level.  The negotiation may be for many or few
resources.  Once a Researcher redeems a ticket for resources at the
AM, the AM exposes some API to the researcher to allow it to execute
privileged operations (e.g. load a custom kernel, if some deep control
is needed) or just ssh to them or similar if only superficial control
is needed.    Whether researchers are well-known or anonymous is up to
the Clearinghouse and how it authorizes/identifies users (in
conjunction with its AMs), which is independent of the transfer
sequence above.  Access Control seems to be a configuration attribute
independent of the resources which an AM and Researcher could possibly
negotiate separately, while QoS could be one particular attribute of a
resource negotiated between Clearinghouse and Researcher.

-David


GENI Control Framework WG Members, and Dave Irwin of UMass Amherst,


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:  5)  Transactions

I would like to invite Dave Irwin of UMass Amherst to begin the
discussion, and then others are invited to join in.



6a)  On the call, Harry indicated that the basic transaction that the
CF needs to support is that researchers can obtain resources, and then
control them to execute experiments.

Dave indicated that ORCA is based on two tenets:  resources are
shared, but not forever.

The group then had a wide-ranging discussion on the nature of
transactions, and felt that the CF needs to support a range of
transactions.


6b)  The current DRAFT of the CF Requirements document covers a
variety of situations in Section 5.5.4.

But, this section is a somewhat confusing listing of options, without
a clear structure.


6c)  Perhaps we should instead consider this approach:

       The CF allows the authorization and assignment of resources
from aggregates to researchers...

       through negotiation of a "lease" (or "ticket", but currently
ticket implies a certificate) that includes:

               characteristics and/or identity of resource

               time (can be now, specific, or indefinite) and
duration (must be limited)

               level of assurance

       with ability for aggregate to change negotiated "lease"

       with ability for researcher to change negotiated "lease"




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