Policy Committee Meeting Summary:
September 13-14, 1999
Arlington, Virginia
Participants
Members |
Representatives |
John Merrill (Chair)
Ken Crawford
Robert Fox
David Knight
James Moore
Charles Murphy
Mohan Ramamurthy
Julie Winkler
UPC Staff
Sally Bates
Ben Domenico
Joanne Graham
Jo Hansen
Linda Miller
Russ Rew
|
David Carlson (UCAR/ATD)
Harry Edmon (ATAC)
Jack Fellows (UCAR/UOP)
David Fulker (UPC)
Mary Glackin (NOAA)
Bernard Grant (NSF/ATM)
Clifford Jacobs (NSF/ATM)
Jennie Moody (Users Com.)
Tim Spangler (UCAR/COMET)
George Serafino (NASA)
Others
Susan Jesuroga (UCAR/COMET)
|
Administrative Matters
- The Committee welcomed Mohan Ramamurthy as a new member
and thanked Jim Moore and Ken Crawford for their contributions.
New member Mike Biggerstaff was unable to attend (due to field
program involvement) and expects to participate in the next
meeting. Charlie Murphy is the new Policy Committee liaison
to the Users Committee.
- Correction to last meeting's summary: Under International Data Issues, the
summary should reflect that Knight's response to MetEirann's objections was
to block access to his data by any except .edu sites.
- The next meetings of the Policy Committee will be:
- January 27-28, 2000 in Boulder
- May 22-23, 2000, possibly at SSEC in Madison, Wisconsin
Status Reports
Director's Report
Copies of Fulker's status report and Graham's budget report were distributed
at the meeting. Fulker's assessment was that there were few problems in most
areas. Caution is warranted in three areas: User Relations (because some users
have expressed a preference for greater emphasis on legacy applications and
on AWIPS), Data Distribution (because the rapid growth in data distributed via
the IDD is straining the system), and in Finances (because projections indicate
a depletion of reserves in the future, exacerbated if any advancement projects
are attempted).
Discussion
- There was a great deal of discussion on the scalability
of the IDD and the reliability of the LDM. More data are being
queued for longer periods and data over an hour old are lost.
At the same time, the amount of data being pushed is increasing
rapidly. Some of the limitations of the LDM are: queuing is
limited to one hour and is limited in size to 2 GB; it supports
a very limited number of data feed types; changing an LDM's
data source is handled manually (except where individual universities
have implemented ad hoc fail-over scripts), and hence the
IDD topology is essentially static and it is maintained through
human effort (even though automation could yield better results).
- The ability to prioritze data was discussed. The UPC doesn't track data
utilization. Filters are already employed at leaf nodes; however, relay nodes
may not employ filters because they must relay whole data streams.
- It was suggested that the ATAC might want to discuss IDD
scaling. It was also suggested that scaling the IDD represents
a fundamental computer-science problem in networking technology.
Jacobs suggested that the problems be identified and a presentation
on this be made to CISE as groundwork for a proposal to CISE
for additional funds to help address the problems.
- Graham noted that she has established a Web page (accessible from the Unidata
home page) for organizing next summer's Users workshop
Users Committee Report
Moody noted that the Users Committee had not met since the
last Policy Committee meeting. The next meeting will be at the
end of September and will focus primarily on next summer's workshop.
Moody reported that she's asked the committee via e-mail for
comments on interest in and access to satellite data. (A copy
of her e-mail is in the notebook.) One respondent suggested
replacing McIDAS satellite data with NOAAport GINI imagery;
she answered by observing that the GINI images (as a result
of remapping, etc.) are unsuitable in many research contexts.
Murphy's status report on MetApps is in the notebook.
Discussion
- Someone suggested carrying both GINI and McIDAS satellite information on
the IDD. It was noted that this would add significantly to the IDD burden.
- Fox noted that SSEC is willing to change the contents of
its data stream. It could easily add more data without additional
costs; the limits are not contractual or technical, but result
from Unidata decisions on data-stream contents. He also noted
that SSEC's ADDE server allows another approach for accessing
satellite data. ADDE, however, is limited to those in the
community who use McIDAS.
- Fox reported that GOES-9 and -10 data will be rescued to
3590 tapes by the end of this calendar year, and the rescue
of GOES-8 data should be completed by the end of March 2000.
The completion of the GOES-8 rescue means that all GVAR data,
which spanned the timeframe from 1994 to the current date,
will be available on 3590 tapes for cheaper and more rapid
access. Work will then commence on rescuing GOES modes A,
AA, and AAA data (GOES data prior to GVAR) from 1978 to 1996.
- Concerning MetApps, it was noted that, despite the addition of new members,
the number of people actively involved in discussions hadn't changed. The
idea of rotating membership was mentioned. Murphy noted that UMADA was working
well and had been used successfully over the Spring and Summer 1999, but Knight
disagreed, noting that although UMADA delivers all comments as mail to the
participants, participants must reply using the Web-based system, which is
cumbersome and limits discussion. Overall, however, the MetApps/developer
arrangement seems to be fruitful. Rew reported that the developers are happy
with the input they're receiviing.
NOAA Report
NOAA Report Glackin report that all AWIPS systems have now
been deployed and the agency is in the process of commissioning
them. The NWS has continued to close older, smaller offices
as planned. Only 22 remain to be closed. In the process of affecting
these changes, some interesting developments have occurred.
For the Willeston, ND, office the NWS will be adding a commercial
S-band doppler radar. For the Erie, PA office, the NWS will
use the FAA radar. It will be an ASR-11 to be installed late
in 2001. Data from the new commercial S-band doppler will be
added to the NOAAport data stream.
NOAA has also released its strategic plan, available on line at: www.nws.noaa.gov/sp/
This plan recognizes the provision of climate services as part of the NWS mandate
and emphasizes science and technology infusion in the form of USWRP, community-based
modeling efforts, and the C-STAR program.
The latest information concerning NIDS is in the notebook; the agency is testing
its central server and working to produce radar mosaics and on adding radar
data to NOAAport.
In other areas, NWS is addressing how to handle the projected increases in
model data volumes. One approach may be to divide the model data into geographical
sectors similar to what is being done on NOAAPORT now with GOES East and GOES
West. The agency is also trying to establish standards for decoders.
Discussion
- The schedule for adding radar mosaics to NOAAport depends
on the NWS being ready to do so, not on agreements with the
NIDS vendors.
- NWS plans to push as well as broadcast their data; they're testing purchased
software (not the LDM) to push data; they are increasing the bandwidth allotted
for this.
- Unidata's agreement with WSI to distribute NIDS data is renewed annually.
NASA Report
Serafino reported on
- The Status of EOS
- Landsat has been launched and data are available
- QuikSCAT has been launched; data will be available early next year.
- Terra is now scheduled to be launched next month. Instruments aboard
include MODIS, MISR, ASTER, MOPITT.
- 1-degree assimilation test flow of EOS data will begin next month
- EOS PM-1 is still scheduledfor launch at the end of next year
- End-to-end testing continues on EOSDIS Core System (ECS) at the DAACs
- Access to data will be through DAACs, ESIP (Earth Science Information
Partners), RESACs (Regional Earth Science Application Centers), remote
sensor application data centers, and others; there is no policy of charging
for EOS data.
- Data Access Issues
- Goal: making data holdings interoperable at several levels (users applications,
systems applications, data usage, acquisition, and searching)
- Researching mechanisms for reducing volumes to maximize information
content (content-based searching with coincident data searching); goal
is to reduce/mine data before distribution or to reduce/mine data within
the processing stream (e.g., by obtaining algorithms from users)
- Employing "Beowulf Cluster"--PC-based parallel processsing cluster.
NSF Report
Jacobs reported that NSF is engaged in searching for replacements for GEO and
ATM directors and that there is no news on the NSF budget. Grant reported that
8 equipment grant proposals were funded (19 proposals had been received). Since
the panel only awarded $77,000, the remaining monies will be carried over into
the next year. He noted that the review panel found an unusual number of the
proposals to be substandard.
Discussion
- The equipment-grant system isn't broken, the community simply needs to help
sites develop appropriate proposals. A buddy-system was suggested, as was
phone calls to sites.
COMET Overview
Spangler and Jesuroga gave an overview of the COMET program.
Discussion
- COMET is moving to AWIPS and doesn't have the resources to support both
GEMPAK and AWIPS. They are considering moving case studies to the lowest common
denominator format (e.g., GRIB). The next case study, however, will be in
both GEMPAK and AWIPS formats
- COMET was commended for releasing the case studies and for developing GARP.
Both have been most welcomed by the Unidata community.
- Community contributions may provide a mechanism for continuing the creation
of GARP case studies.
GEMPAK and AWIPS
Fulker noted that the letter from Gen. Kelly on GEMPAK (in the notebook) affirms
that the NWS is ceasing development of GEMPAK and eliminates any doubt about
UPC flexibility to use the GEMPAK code as it wishes. This means that McIDAS
and GEMPAK are now equally available to Unidata participants without the need
of licensing. The letter does indicate that a few enhancements to GEMPAK will
be forthcoming before development ceases entirely. The letter also affirms that
the NWS is interested in cooperating with universities. Fulker believes that
cooperation may now be in the realm of decoder development.
For some users, the implications of this letter cast doubt on the viability
of GEMPAK and GARP in Unidata, considering our prior reliance on NWS and COMET
staff for advances. In contrast, Fulker believes that the Unidata staffing in
this arena will yield continuing advances (as already demonstrated, e.g., in
adaptations of GEMPAK to utilize new data sources) though the pace will slow.
Fulker asked the committee to consider this matter from a policy perspective:
Is it appropriate to continue the present priorities, which sacrifice the pace
of advancement in legacy software (GEMPAK and McIDAS) so as to achieve more
quickly the benefits of Java-based MetApps software? A closely related policy
question is whether Unidata should directly exploit other software from NWS
and FSL, such as the anticipated Linux version of the AWIPS workstation. Fulker
advocates maintaining the current plan and priorities, and he said the benefits
of the Java/MetApps strategy include: high levels of platform independence;
a component architecture (which allows users to "get under the hood..."); use
of abstract data and display models; less emphasis (than in AWIPS, e.g.) on
powerful data-base servers; and greater potential for the direct integration
of Unidata software with other educational materials that are likely to emerge
as universities exploit advancing technology.
Discussion
- Moving GEMPAK into maintenance only is a real concern to
a number of users. This is balanced by the fact the the Users
Committee reaction to AWIPS was relatively cool
- The focus of universities is to teach the concepts of meteorology.
Since most students will never work in the Weather Service,
some felt that the loss of shared applications may not be
a problem for many campuses.
- AWIPS software is available to universities but is dependent
on platforms (HP workstations) that are beyond the means of
most universities. It also represents the struggle of universities
between being cutting edge or looking like a Forecast Office.
Universities and FOs have different missions but both want
to continue having at least some commonalities. FSL and NWS
are considering a port of AWIPS to Linux, which may make AWIPS
more available to universities. However, AWIPS does not fit
the Unidata model: it's a turnkey system that is not user
configurable.
- The GEMPAK situation exemplifies UPC vulnerability to losing
support for packages developed by others: Vietor's support
for WXP essentially has disappeared, Mary desJardin first
moved from NASA to NOAA and now will focus on AWIPS. SSEC
also has suffered a significant reduction in resources for
software development. This reflects a more general trend away
from software development toward programmatic foci, often
accompanied by the use of commercial packages such as IDL.
This is true at NCAR as well: ATD has dropped development
of Zebra.
- Groups within NCAR use SCD's visualization tools and IDL.
IDL is widely used in part because it can be used to analyze
netCDF files. Hence Unidata's efforts have actually facilitate
the use of IDL, and this underlines the importance of well-designed
interfaces, built on effective, useful abstractions. Separating
software levels appropriately through such interfaces not
only allows more collaborative work, but also ensures that
others WILL write applications.
- Without shared applications, there is no Unidata community.
- The committee agreed that retaining access to COMET case
studies is a major concern. UPC funding for case study support
is level, but the UPC is committed to ensuring accessibility
by universities.
- The loss of GEMPAK development may not become a problem
for some time. Some noted that universities are slow to maximize
their use of applications capabilities; by the time most fully
utilize GEMPAK to the point of desiring additional capabilities,
MetApps may be operational. Others were concerned that MetApps
development may take much longer than this anticipated lag
time.
- There was disagreement on whether community contributions
would really appear. Experience indicates that universities
won't step in to do development, that development needs to
be centralized. This was countered by the observation that
community members are actively involved in building on VisAD.
- It was suggested that if MetApps development is not proceeding
as rapidly as predicted, this needs to be identified early
on and additional resources found to aid in the development
work.
Participation Policy
Fulker explained that the purpose of the draft document (distributed
by e-mail) was to codify who it is the UPC should support. The
method currently used to track users (the software licensing
method) is out of date.
Discussion
- One purpose of a new participation policy is to provide guidelines to UPC
staff on how to respond to queries from people ouside of Region 4. The desire
to document participation is also driven by the need to account for resource
use to the NSF (to measure success).
- UPC needs to decide what information it needs/wants. For some software
(netCDF), there is no way of determining who is "participating" beyond simply
tracking downloads.
- The committee warned the UPC that there may not be any
way to design a fully automated system. This is particularly
true when data are involved. Does downloading software allow
you to subscribe to data?
- Virtual universities pose a risk of being unable to identify who should
be supported--there may be no single representative. Fulker noted that Unidata
currently encourages and supports one feed per university, so funneling support
should not be a problem
- If you create a database of users, then you must be prepared to open the
database to individuals under the Freedom of Information Act.
- Access software, access to support, and access to data may be separate
issues. Does participation include those who access all UPC services? In terms
of resources, the crux of defining participation is who is able to request
support. Automation of the LDM would result in the topology of the IDD being
self-determined to a large extent.
- Unidata applications software is not in the public domain.
GARP is copyrighted by COMET; McIDAS is copyrighted by SSEC
and the UPC is allowed to distribute McIDAS to users for educational
and research uses only. These restrictions must continue to
flow through to users. COMET uses "point-and-click" licenses
that advise users of the existence of the copyright and the
restrictions on use.
Action 1:
The UPC will develop a prototype user tracking information system and report
the status of this effort at the next meeting.
Action 2:
Fulker will refine the draft Participation Policy and this
will be an agenda topic at the next meeting. The refinement
needs to strengthen the section on points of contact in respect
to data flows and include more on Unidata's core constituency.
Action 3:
Merrill will write an article for the next issue of the newsletter
on the increase in data usage, the potential future difficulties
these pose for the LDM, and staff efforts to meet these challenges.
The article will include information on possible solutions.
Strategic Implications
Fulker noted that the level of growth in the IDD is leading the UPC to questions
its current priorities regarding LDM development. The UPC had assumed that data
flows would be limited by Internet bandwidth, but this has become dfficult to
identify (e.g., the Abilene project's interest in CONDUIT). Furthermore, the
SSEC no longer links the size of the McIDAS stream to a specific costs.
Fulker noted some possible approaches:
- (less desirable) Assign data priorities, with guidance
from Users Committee.
- (more desirable) Develop mechanisms such that local decisions
can be made in the context of local capacity. This probably
requires creating MANY data feeds.
- Nearly any solution will require LDM development, including
automated routing, larger queues, a simpler LDM-administration
interface (perhaps in conjunction with reimplementation in
Java), and the ability to handle more feeds.
Discussion
- It is not clear whether the transfer of NIDS to NOAAport will create a
problem for the LDM. The importance of success is illustrated at Oklahoma,
where NIDS are the most requested data by at least one order of magnitude.
One indicator suggests that no problem will arise: NCAR/RAP currently handles
NIDS from all radars with a single LDM.
- Policies on data use are easier to set. Restricted data are distributed
separately. But the rapid increase in the amount of available data results
in two problems: bandwidth limitations and mechanisms for easy selection by
users of portions of the stream.
- Millersville is already acting as an intelligent filter for some satellite
data. In addition, the UPC might consider developing a simple application
to allow users to select radar/satellite/model/sounder/... by geograpical
area.
- It was very clear in the context of all the discussions that the Policy
Committee felt that the current priorities vis a vis support for legacy applications
and MetApps development is appropriate.
Action 4:
LDM development will be a topic on the agenda for the next meeting.
Action 5:
The UPC will inform the community of the status of GARP and GEMPAK.
Action 6:
Data use policies will be a topic on the agenda for the next meeting.
Action 7:
Fulker will draft a policy on the distribution of NOAAport data by Unidata..
Action 8:
In consultation with SSEC, Unidata will draft a software distribution approach,
based on shrink-wrap licensing, for Policy Committee consideration in January.
Action 9:
An overview of the SuomiNet proposal will be on the agenda for the next meeting.
Unidata Homepage
This page was Webified by Sally
Bates.
Questions or comments can be sent to <sally@unidata.ucar.edu>.
This page was updated on .