[glue-wg] TeraGrid capability registration in information services

[JP Navarro]

Greetings,

During Q2 2007 the TeraGrid will deploy new information services that
will publish which high-level capabilities are available on individual
TeraGrid resources, as well as information about the software and
services providing each capability. The new information services are
part of our next generation Coordinated TeraGrid Software and Services
"CTSS" version 4 project.

The reason for this e-mail is to share aspects of our information  
service
plans, like a generalized capability publishing entity that can  
encompass
more than computing and storage elements, and to solicit your feedback.

Would any of the material below provide valuable GLUE 2.0 use cases?  
Should
it be posted in a TeraGrid document in the GLUE 2.0 Requirements area?

A summary of our capability, service, and software information follows.

=======================
Examples of TeraGrid capabilities:
   Remote compute (i.e. Grid job submission)
   Data transfer (i.e. Grid file transfer)
   Data management
   Remote login (i.e. Grid login)
   Parallel application support (i.e. MPI)

Initially, we plan to publish the following capability attributes:
   SiteID                  Globally unique site ID offering the  
capability
   ResourceID              Globally unique resource ID/cluster  
offering the capability
   UniqueID                Globally unique capability ID
   Name                    Capability (kit) name
   Version                 Capability (kit) version
   supportGoal             Targeted support level
   supportLevel            Current support level
   statusURL               URL to a monitoring status page
   userSupportOrganizaton  Descriptive name for the user support  
organization
   userSupportContact      How to contact the support organization

Sample XML:
<kitRegistration UniqueID="dtf.uc.teragrid.org"  
timestamp="2007-02-01T04:37:10Z">
    <SiteID>uc.teragrid.org</SiteID>
    <ResourceID>dtf.uc.teragrid.org</ResourceID>
    <kit UniqueID="RemoteCompute.dtf.uc.teragrid.org">
       <Name>RemoteCompute</Name>
       <Version>1.0</Version>
       <supportGoal>production</supportGoal>
       <supportLevel>testing</supportLevel>
       <statusURL>http://someurl.teragrid.org/resource/kit/</statusURL>
       <userSupportOrganization>TeraGrid</userSupportOrganization>
       <userSupportContact>mailto:help at teragrid.org</userSupportContact>
    </kit>
    <kit UniqueID=DataTransfer.dft.uc.teragrid.org">
       ...
    </kit>
</kitRegistration>


Service registration
====================
Examples of TeraGrid services:
   GT 4 Pre-WS GRAM service
   GT 4 WS GRAM service
   GT 4 MDS service
   GT 4 GridFTP service
   gsiSSH service
   SRB service

Initially, we plan to publish the following service attributes:
   SiteID                  Globally unique site ID offering the service
   ResourceID              Globally unique resource ID/cluster  
hosting the service
   UniqueID                Globally unique service ID
   Name                    Human friendly service name
   Type                    Service type
   Version                 Version of service type
   Endpoint                Service contact point


Software registration
=====================
Examples of TeraGrid services:
   Condor 6.9.1
   mpich-gm 1.2.6e build with Intel 8.1
   GT 4 MDS service

Initially, we plan to publish the following software attributes:
   SiteID                  Globally unique site ID offering the software
   ResourceID              Globally unique resource ID/cluster  
hosting the software
   UniqueID                Globally unique software ID
   Name                    Software name
   Version                 Version of software
   HandleType              Method for accessing software from Unix shell
   HandleKey               Argument/key for the method
   Default                 Available in default login (and job)  
environment


There is some obvious overlap between service and software  
attributes. Services
are available to users thru an Endpoint, software is available to  
users thru some
Unix level access method (handle type and key).  Services and  
software would be
associated with one or more capability kits.

Of these 3 types of information, service is the only one that  
currently exists
in GLUE, I believe. We selected a subset of the GLUE 1.3 Service  
attributes that
the TeraGrid has a use case for.

We are also considering publishing Site information, and would choose  
a subset
of GLUE Site attributes. If our users demanded additional compute  
element and
storage element information, we would also adopt subsets of these  
entities
from GLUE.

Regards,

JP Navarro
------------------------------------------------------------------------ 
------
John-Paul Navarro                                               (630)  
252-1233
navarro at mcs.anl.gov                            http://www.mcs.anl.gov/ 
~navarro
TeraGrid Software Integration                Univ of Chicago/Argonne  
Nat. Lab.
GPG: 4EA9 C86B C0F0 113D 6306  98B7 3649 D6CB EFA8 4133