Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.adass.org/adass/proceedings/adass99/P3-14/
Дата изменения: Sat Oct 7 00:00:54 2000
Дата индексирования: Tue Oct 2 05:11:54 2012
Кодировка:
Поисковые слова: п п п п п п п п п п п п п п п п п п п п п п
|
Next: The BeppoSAX Low-Energy Concentrator Spectrometer Archive at the Space Science Department of ESA
Up: Archiving
Previous: The Multimission Archive at the Space Telescope Science Institute
Table of Contents -
Subject Index -
Author Index -
PS reprint -
Wieprecht, E., Wiezorrek, E., Bauer, O. H., Huygen, R., Vandenbussche, B., Claes, P., & Mathieu, J. J. 2000, in ASP Conf. Ser., Vol. 216, Astronomical Data
Analysis Software and Systems IX, eds. N. Manset, C. Veillet, D. Crabtree (San Francisco: ASP), 172
Object Oriented Near Real Time Packet Distribution as Part of the FIRST
Integrated Network and Archive System
E. Wieprecht, E. Wiezorrek, O. H. Bauer
Max Planck Instititut fuer extraterrestrische Physik, Garching, Germany
R. Huygen, B. Vandenbussche
Institute of Astronomy K. U. Leuven, Leuven, Belgium
P. Claes, J. J. Mathieu
ESA - Astrophysics Division, Noordwijk, Netherlands
Abstract:
The Far InfraRed and Submillimetre Telescope (FIRST) is the fourth cornerstone
mission in the original European Space Agency's HORIZON 2000 program.
The Integrated Data Archive System
(FINDAS) will be the base of the
distributed ground segment.
This paper shows aspects of the development of a prototype
application to test the feasibility of near real time distribution
of Telemetry Packets to the different instrument control centers.
FINDAS is based on a 3-Tier Architecture.
We discuss the development of the middleware component (an object
server) and front end (a real time client application).
FIRST will be a 3.5 m telescope, covering the wavelength range of 80 to 670
m and hosting three scientific instruments: the heterodyne
instrument (HIFI), the photoconductor instrument (PACS) and the bolometer
instrument (SPIRE).
Currently the launch is foreseen in early 2007. The satellite will orbit the
second Lagrangian Point (L2) and will operate for about 4.5 years.
The FIRST ground segment consists of five major parts: The Mission
Operations Center (MOC) for operating the satellite, the FIRST Science Center
(FSC) for coordination of overall science activity and the three Instrument
Control Centers (ICCs) for operation and characterization of the instruments.
For FIRST a novel geographically distributed ground segment is envisaged.
Mission operations center, science center and instrument teams will not be at
the same location. Goal is to reduce operational costs by avoiding traveling of
experts for various aspects of the mission.
The FIRST ground segment will be based on the object oriented FIRST Integrated
Network and Data Archive System - FINDAS. It will manage all information
relevant to the FIRST project throughout all phases of the mission, from
spacecraft/instrument development, through all tests, checkout, operations and
post mission phase inclusive archiving. FINDAS is not only the repository for
all FIRST related information, it also supports:
- Flexible, efficient and controlled access to up-to-date FIRST
information, data and software for all ground segment elements,
- Homogeneous configuration management for all information, data and
software,
- Homogeneous data analysis environment for development and operations for
the different ground segment elements,
- Homogeneous data access and analysis environment for observers.
The architecture of the operational environment will be kept similar throughout
all phases of the development.
Some elements in the FIRST ground segment will not be available in the early
phases like ground tests on the integrated instrument level. Their places will
be taken by mock-ups, pieces of software which perform a subset of their tasks.
As test procedures require permanent online monitoring it is essential to have
a near real time data access. This allows immediate recognition of abnormal
behavior of the instrument and instantaneous reactions in the case of critical
situations. Also ``manual commanding'' and execution of the command will be
verified in real time. Furthermore the instantaneous data access allows
efficient data analysis and test cycle optimization.
Goal of the prototype is to provide all players in the ground segment with the
possibility to gain experience with a distributed object oriented database
system, to get a feeling of the technological feasibility of the concept and
to refine the user requirements of the whole system.
A real time application was chosen in order to demonstrate the technological
feasibility of displaying real time data from FINDAS during ground based
instrument tests and during certain phases of the mission rather than
listening for telemetry packets ``online''.
A 3-tier architecture approach was chosen. The user interface is the TMViewer.
The TMViewer communicates with an object server. The object server is extracting
telemetry packets from the object database and is hiding the physical and
logical structure of the database for the client application (TMViewer).
The TMViewer client should prove that a system build on this architecture can
provide us with packets within the time frame we envisage, one 65 kB TMPacket
every two seconds up to eight 4 kB packets every single second.
In order to verify this, the following issues are tested while retrieving
telemetry from the FINDAS prototype: the integrity of each packet, the order
of retrieval and missing packets, the performance under different
circumstances (turn around time between instrument status change and update
of the view in the TMViewer), to connect to database with multiple clients,
different data rates (200 Kbits/sec to 4MBits/sec), packet sizes
(4 kB to 65 kB) and packet modes, has the size of the DB or complexity
of the data model an influence on the performance?
For remote object access, necessary for the communication between the three
players in this architecture, several possibilities were explored:
- Simple socket programming which is in general very effective - but not
well suited for complex data types,
- Simple Remote Procedure Calls (RPCs) which is based on sockets, however
mostly easier,
- Distributed Component Object Model (DCOM) from Microsoft - which is
operating system dependent,
- Common Object Request Broker Architecture (CORBA) which is platform
independent and support JAVA and C++. But the implementation needs some effort,
- Remote Method Invocation (RMI) - which is platform independent and easy
to implement. But client/server applications have to be written in JAVA.
For the prototype of the real time TMViewer the combination C++ / CORBA and
JAVA/RMI have been tested. We used the ORBIX CORBA from IONA (version
2.7/3.0) which is supporting CORBA (version2.0/2.1) and JAVA version 1.2.
A difficult trade-off to make is the eagerness of the different components.
The object server could be constantly ``polling'', i.e. query for the next
object, if not available wait and try again. Some simple intelligence can be
built in to optimize the frequency of issuing a new query based on the
frequency of success. Another possibility is to use notification/signal
mechanisms of modern OO databases triggering a signal upon the creation of a
new instance of class.
The latter solution might be better to avoid unnecessary load on the
database,
but the first solution might result in a better throughput and response time if
the data rate is close to the capacity of the system.
The development of applications to test the FINDAS prototype is also used to
learn how to co-operate with several groups writing applications that work on
objects in a single data model.
For the design of the data-model, UML diagrams are taken as ``communication
language'' between the different parties.
Figure 1:
UML: The prototype data model.
![\begin{figure}
\epsscale{0.7}
\plotone{P3-14a.eps}
\end{figure}](img2.gif) |
The proposal handling system builds the front end to the test
observer. It creates and maintains proposals and proposal objects.
The MOC/FSC simulator converts the proposal to an observation.
The TMGenerator is analyzing the observation and creates the
TMPackets. The Real Time Object Server polls the DB for new
TMPackets. If a new TMPacket is available it is extracted and the
information is displayed by the TMViewer, the interface to the test
person.
All the components of this scheme are build by different parties. The data
model is the base for the developers. See Figure 1.
We were able to write a simple application in a 3-Tier architecture in
four man weeks (inclusive learning simple basics about remote object
access and OODB).The exercise was rather helpful for defining the User
Requirements on the whole system. UML communication was essential for
the distributed s/w development process. A software tool supporting UML
(e.g. Rational Rose) proved to be useful.The C++/CORBA requires more
learning/understanding then JAVA/RMI. We were troubled by
ORBIX2.7/CORBA2.0 - ORBIX3.0/ CORBA2.1 inconsistencies. Until now, we
only tested the polling mechanism. Further tests, based on this
configuration, will show us whether the technology used is suitable to
serve as backbone for a scientific satellite ground segment.
Acknowledgments
Ekkehard Wieprecht acknowledges financial support from the Deutsche
Forschungsgemeinschaft (DFG).
© Copyright 2000 Astronomical Society of the Pacific, 390 Ashton Avenue, San Francisco, California 94112, USA
Next: The BeppoSAX Low-Energy Concentrator Spectrometer Archive at the Space Science Department of ESA
Up: Archiving
Previous: The Multimission Archive at the Space Telescope Science Institute
Table of Contents -
Subject Index -
Author Index -
PS reprint -
adass@cfht.hawaii.edu