Skip to Content.
Sympa Menu

perfsonar-dev - Definition of perfSONAR

Subject: perfsonar development work

List archive

Definition of perfSONAR


Chronological Thread 
  • From: "Eric L. Boyd" <>
  • To:
  • Subject: Definition of perfSONAR
  • Date: Mon, 09 Oct 2006 17:23:12 -0400

Greetings,

As discussed in today's perfSONAR development meeting, we propose to change/refine our definition of "perfSONAR". The goal of these changes is to:

a) Enfranchise more (all?) ongoing perfSONAR-related efforts as "perfSONAR".
b) Allow us to present a coherent answer to "what is perfSONAR" to different communities.
c) Reflect the reality of unacknowledged, real processes in the perfSONAR consortium and incorporate them into the perfSONAR project.

We do NOT intend to reduce the incentive to collaborate with each other or work together on common projects.

Please review these changes and we will revisit the topic first thing in the morning, review the proposal, make changes if necessary, and jassign action items.

Changes:

1) There will no longer be "perfSONAR core services" or "perfSONAR affiliated services" or "perfSONAR unaffiliated services".

2) We define four new classes of perfSONAR-related activity

a) perfSONAR proposed services
b) perfSONAR services
c) perfSONAR clients
d) perfSONAR suites

3) PerfSONAR proposed services are defined as:

a) Conforming to the syntax (binary representation of the message) and semantics (business logic) of the protocol standard
ACTION ITEM: Define the syntax (largely done via the schema).
ACTION ITEM: Define the semantics (largely incomplete)

b) Having a base level of documentation
ACTION ITEM: Define base level of documentation

c) Conforming to the naming scheme for services
ACTION ITEM: Define naming scheme for services. Consider including developing institution, development language, service type, tool, e.g. DFN-PERL-MA-BWCTL service.

d) Having a non-coercive open source license.
ACTION ITEM: Determine a non-exhaustive list of likely/preferred licenses (e.g. Apache, modified Berkeley, etc.)

Note that there are no requirements about language of implementation, installation, packaging, release management, documentation format, etc.

4) PerfSONAR services are defined as:

a) Meeting the aforementioned requirements for perfSONAR proposed services.

b) Having successfully concluded an interoperability bake-off.
ACTION ITEM: Define when a bake-off can occur and what is considered "success". An interoperability bake-off occurs when there exist 2 or more services with a common interface and one or more services or clients interoperating with those services. Bake-offs can happen repeatedly, if there is a need. The goal of these bake-offs is to validate assumptions made about unspecified syntax or semantics and incorporate such assumptions into the protocol standard.

Note there is no suggestion that perfSONAR proposed services are of lesser value than perfSONAR services.

In addition, we have define a bunch of processes for the various aspects of these definitions, but we didn't record these very well, so I'll need my memory refreshed in the morning.

5) PerfSONAR clients are defined as:

ACTION ITEM: Come up with a definition as we haven't covered this topic yet.

6) PerfSONAR suites are defined as:

a) A set of 1 or more proposed services, services, and clients
b) Issued by an organization (e.g. GEANT2 JRA1, Internet2, the perfSONAR consortium, etc.)
c) Recommended for a collection of adopters (e.g. European NRENs, LHC community, generic adopters, etc.)

A perfSONAR suite may have additional requirements of included services (e.g. a common installation process, all written in Java, etc.) Various perfSONAR suites may overlap or not with each other.

To reiterate, it is still in our common best interest to work towards common suites, but it is not a requirement.

Comments and discussion welcome.

--Eric

--
Eric Boyd
Director of Performance Architectures and Technologies
Internet2
1000 Oakbrook Drive, Suite 300
Ann Arbor, MI 48104 USA
(734)352-7032

The Fall 2006 Internet2 Member Meeting:
Because the best way to predict the future is to keep inventing it.
http://events.internet2.edu/2006/fall-mm/



Archive powered by MHonArc 2.6.16.

Top of Page