Hi Stephen and Jeff,

who is "in charge"/"can change" the implementation of the algorithm for 
the EstimatedTraversalTime (now GlueCEStateResponseTime) is EDG WP4, 
especially Thomas Roeblitz (roeblitz@zib.de) and Jörg Meltzer 
(meltzer@zib.de).

You can gather the latest implementation here:
http://datagrid.in2p3.fr/cvsweb/fabric_mgt/resource_mgt/information_providers/ce-all.pl?only_with_tag=HEAD

I'm actually not aware of the current algorithm they use, since we are 
still testing fixing other problems (e.g. complexity of LSF queue 
configuration). Anyway I understand this is an important issue since it 
can affect the broker decision since the mentioned attribute is the 
default one to rank CE's.

Within Glue we have not specified any algorithm, since it is out of our 
scope. I suggest you to write/interact with EDG-WP4 people.

For any other glue schema info, you can write me. The resource page is 
this one:

http://www.cnaf.infn.it/~sergio/datatag/glue/index.htm

Cheers,

Sergio


Stephen Burke writes:

  > On Wed, 5 Mar 2003, Jeff Templon wrote:

  > >   o JT (and he hopes others as well) will make proposals for 
"better ETT"
  > >     (meaning an algorithm suitable for a ce-info-provider script)
  >
  > Since we will be adopting the Glue schemas and info providers from
  > testbed 2, do you have any idea what the mechanism will be for getting
  > things changed? It's been hard enough with it happening inside EDG!


-- 
Sergio Andreozzi

Istituto Nazionale di Fisica Nucleare - CNAF
Viale Berti Pichat, 6/2
40127 BOLOGNA
ITALY

Tel +39 051 609 2860
Fax +39 051 609 2746

E-mail : sergio.andreozzi@cnaf.infn.it
Website: www.cnaf.infn.it/~sergio