r1 - 26 Sep 2007 - 10:38:23 - RobertGardnerYou are here: TWiki >  Admins Web > SurveyFormSite
Fill out and return to John Weigand (weigand at fnal dot gov).


The site survey we talked about Friday that was sent out some time ago and allowed us to determine the normalization factor used in reporting account data to APEL/WLCG is at the back of this email.

For the OSG Sites that properly filled out a survey of the actual
composition of their farm, we use the weighted average of the
normalization factor.

For the sites that did not provide any information about their farm's
composition, we use the actual normalization factor from a representation farm (CMS Tier1 at Fermilab)

The normalization factor is the value of SpecInt 2000 divided by 1000.

Hope this helps
John Weigand


You're receiving this request for information because you are
listed as the site admin for one or more of the registered OSG
Sites. There is a very urgent need to obtain correct and up to
date site configuration information so that the OSG can report
these metrics to the funding bodies and so that the OSG will
know the computing capacity it provides. OSG is moving to
formal reporting of metrics, both for our own use and because
of  funding agency requirements. We must collect accurate
data and be able to support it's validity.

Because of the high level of detail in GLUE we will not be
making an exhaustive study. The questions below are intended
as a sampling of key data that can also be used to generate
metrics we need for funding agencies and performance reports.

We will use the site name registered with the GOC as we
do our survey. However, we have found this name is not
consistently applied in configuration of the sites.
Please supply the site name that GLUE data
has for your site that is reported with GIP.

We will be contacting the site administrator as reported
by the GOC database. If that is not the correct person,
please let us know the correct name. We will need the
name, phone number and email address.
System Administrator is defined as the person who would
maintain the site and correct problems. This should be a
person who can answer detailed questions about the
configuration of the site.

The GOC does not have this information in its database.
Please enter the name, phone number, and email address
of a person or helpdesk that would answer user questions
about submitting jobs or the operational status of the site.

The physical location of the site. This need only be
organization, city, state, country. Example: Fermilab,
Batavia, Illinois, USA.

Number of CEs at this site:
CE as defined by the GLUE Schema. Some sites may
refer to this as a Head Node or Gatekeeper.
Please explain your definition and describe the
implementation. Sites may or may not have more than
one CE. Multiple CEs may share worker nodes.

Number of SEs at this site:
SE as defined by the GLUE Schema. This is a storage
element and a site may have zero, one or more SEs available.
Please explain your definition of an SE and describe the

We would like to collect detailed information regarding
CEs available to OSG. CEs are usually identified by the hostname.
The format is hostname:port/jobmanager-batchmanager-queue.
Example of GLUE CE ID: rsgrid3.its.uiowa.edu:2119/jobmanager-condor-atlas

CE SubCluster Count:
How many subclusters does this CE manage?
A subcluster is defined as a group of homogeneous
processors managed by a CE.
Each CE may have one or several subclusters.

CE OSG Restrictions:
What policies are enforced for OSG by this CE?
For example: Does the CE limit OSG to a percentage of
the job slots available on this CE?

This is somewhat vaguely defined in GLUE.
Please supply the identity you use for this parameter
when configuring GIP. A subcluster is a group of
homogeneous processors. This is very specific.
Each group is composed of a particular processor
type and model. Thus a group of 100 AMD opteron
1.4 GHz processors would be a different subcluster
than a group of 100 AMD opteron 3 GHz processors or a group
of 100 Intel P4 2 GHz processors.
One of our concerns is that subcluster information is not being correctly
reported by GIP, if reported at all.

The following subcluster information very specifically identifies the
processors that compose the subcluster.
Please use "cat /proc/cpuinfo" to obtain this information.
We will also require the number of processors of each type.

This is the number of processors in a subcluster.
We define a processor as a single processing unit.
Thus if the node is comprised of a quad-core processor
we would consider that 4 physical CPUs for the purpose
of this parameter. 1 core=1 processor.
Do NOT include the effects of hyperthreading in this parameter.

The TOTAL number of processors when the effect of
hyperthreading is included.
A quad-core processor with each core processing 2 threads
would be 4x2=8 logical CPUs. This is only included if
hyperthreading is actually implemented for the subcluster.

The following information describes the OS used for
 this subcluster.
Examples are shown for each parameter.
This information may be obtained using "uname -r"
and "cat /etc/redhat-release"(or the equivalent for your OS).

GlueHostOperatingSystemName: Readhat
GlueHostOperatingSystemRelease: 7.3
GlueHostOperatingSystemVersion: 3

A storage element is usually identified by it's hostname.

SE Available Storage:
What is the available storage for this CE in GigaBytes?

-- RobertGardner - 26 Sep 2007

About This Site

Please note that this site is a content mirror of the BNL US ATLAS TWiki. To edit the content of this page, click the Edit this page button at the top of the page and log in with your US ATLAS computing account name and password.


Powered by TWiki
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback