Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.adass.org/adass/proceedings/adass03/P1-21/
Дата изменения: Tue Aug 17 01:20:11 2004
Дата индексирования: Tue Oct 2 04:56:58 2012
Кодировка:
Поисковые слова: п п п п п п п п п п п п п п п
|
Next: The D4A Digitiser
Up: Surveys &
Large Scale Data Management
Previous: Increasing the Accessibility of Green Bank Telescope Data
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint
Hiriart, R., Valdes, F., Pierfederici, F., Smith, C., & Miller, M. 2003, in ASP Conf. Ser., Vol. 314 Astronomical Data
Analysis Software and Systems XIII, eds. F. Ochsenbein, M. Allen, & D. Egret (San Francisco: ASP), 74
The NOAO Pipeline Data Manager
R. Hiriart, F. Valdes, F. Pierfederici, C. Smith and M. Miller
National Optical Astronomy Observatories, Tucson, AZ 85719
Abstract:
The Data Manager for NOAO Pipeline system is a set of interrelated components
that are being developed to fulfill the pipeline system data needs. It includes:
(1) management of calibration files (flat, bias, bad pixel mask and xtalk calibration data.);
(2) management of the pipeline stages' configuration parameters; and
(3) management of the pipeline processing historic information, for each of the
data products generated by the pipeline.
The Data Manager components uses a distributed,
CORBA based architecture, providing a
flexible and extensible object oriented framework,
capable of accommodating the present
and future pipeline data requirements.
The Data Manager communicates with the pipeline modules, with internal and external
databases, and with other NOAO systems such as the NOAO Archive and the NOAO Data
Transport System.
The NOAO Pipeline is a parallel and distributed system that
processes observations from the two NOAO 8K-square CCD wide-field mosaic
imagers in near real time. It performs basic CCD reduction, instrumental
features removal, astrometric calibration, and zero point
calibration (Pierfederici 2004; Valdes, Miller, this conference).
The core of system is composed of several processing modules, which are connected to
form sub-pipelines.
The
OPUS1system is used to control the
process execution environment for each of the modules.
The Data Manager is a subsystem of the NOAO Pipeline that
provides several data management services to the system. These services
include storage and retrieval of calibration files, management of pipeline
module configuration parameters, and recording of processing information.
The extensibility of the Data Manager has been an important goal
in the design, in order to facilitate integration of new services as the
pipeline system evolves and new requirements arise. Therefore, the Data Manager
contains several components implemented as distributed CORBA objects.
Figure 1 presents the overall architecture of the Data Manager.
The system is structured in three layers: the Interface layer, the
Service layer, and the Resource layer. The Interface layer
exposes the Data Manager services to the clients by calling the appropriate component(s) in
the Service layer upon receiving a request. The Service layer, on the other hand,
uses the components in the Resource Layer to access a number of internal services, such
as persistent data in relational databases, and file storage in multiple repositories.
The Interface layer
is composed of the following subsystems:
- DMSocketRouter: Receives requests from
the pipeline modules through a
lightweight socket based protocol. Request processing is asynchronous and concurrent,
which enables the handling of a large number of connections and achieve good
performance.
- DMOpusBlackboards: An experimental
interface, designed to exploit the benefits of the blackboard architecture and components of OPUS,
to integrate the Data Manager with the rest of the system. This feature provides
another method of communication with pipeline modules.
- DMAdministration: Exposes interfaces
for administration GUIs used by system
operators.
The subsystems in the Service layer implements the ``business logic''
of Data Manager operations. The services are based on a rich
object-oriented data model that is maintained in a PostgreSQL database.
The subsystems of this layer are:
- DMCalibrations: Exposes interfaces for the
management of calibration libraries.
- DMParamManager: Maintains
a set of configuration parameters for each pipeline module.
- DMHistorian: Allows the
system to keep a record of all relevant variables needed to review the
pipeline processing history, and to
address provenance issues for pipeline-generated data products.
- DMConfiguration: It maintains a model
with the overall current configuration of the pipeline system.
The pipeline modules can be dynamically chained together at run-time,
using an XML based configuration system. The XML configuration feeds the
DMConfiguration subsystem's data model, and this information is stored
as part of the processing provenance of the generated data products
with each pipeline run.
Figure 1:
Overall architecture of the Data Manager system.
|
The Service layer uses components from the
Resource layer, which is composed by the following subsystems:
- DMFileRepositories Maintains
collections of files stored in multiple repositories, organized in hierarchical directories
on the file system. Maintains a catalog of stored files.
- DMDBManager Provides access to
relational databases. This component is divided in two parts: the interface with the DBMS,
that provides persistence services to objects in the upper layers,
and maintains a pool of database connections; and a set of stored procedures that are used
for processing within the database.
- DMLogger Provides logging services
for all other components of the Data Manager.
- DMExtResources Accesses external
resources such as external catalogs, the NOAO Data Transport system, and the NOAO
Science Archive.
The NOAO Data Manager is being developed as part of the NOAO Pipeline project,
with the goal of providing an extensible, robust, and high performance solution
for the data management needs of the pipeline system, as well as to allow
the integration of other NOAO systems that depends on the data generated in the
pipeline executions, such as the NOAO Archive system.
The NOAO Data Manager is being developed using distributed object middleware
and relational databases. These technologies provide the necessary framework over
which this application can be successfully constructed.
References
Pierfederici, F., Valdes, F., Smith, C., Hiriart, R., & Miller, M.
2004, this volume, 476 .
Footnotes
- ...
OPUS1
-
http://www.stsci.edu/resources/software_hardware/opus
© Copyright 2004 Astronomical Society of the Pacific, 390 Ashton Avenue, San Francisco, California 94112, USA
Next: The D4A Digitiser
Up: Surveys &
Large Scale Data Management
Previous: Increasing the Accessibility of Green Bank Telescope Data
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint