Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.stecf.org/conferences/adass/adassVII/reprints/todyd1.ps.gz
Äàòà èçìåíåíèÿ: Mon Jun 12 18:51:50 2006
Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 03:11:40 2012
Êîäèðîâêà:

Ïîèñêîâûå ñëîâà: earth's atmosphere
Astronomical Data Analysis Software and Systems VII
ASP Conference Series, Vol. 145, 1998
R. Albrecht, R. N. Hook and H. A. Bushouse, e
Ö Copyright 1998 Astronomical Society of the Pacific. All rights reserved.
ds.
The Mosaic Data Capture Agent
Doug Tody and Francisco G. Valdes
IRAF Group, NOAO 1 , PO Box 26732, Tucson, AZ 85726
Abstract. The Mosaic Data Capture Agent (DCA) plays a central role
in connecting the raw data stream from the NOAO CCD Mosaic controller
system to the rest of the data handling system (DHS). It is responsible for
assembling CCD data from multiple readouts into a multiextension FITS
disk file, for supplying a distributed shared image to real­time elements
of the DHS such as the Real­Time Display (RTD), and for feeding data
to the IRAF­based data reduction agent (DRA).
1. Overview
A brief design of the data handling system (DHS) for the NOAO CCD Mosaic
Camera 2 was presented earlier (Tody, 1997). It is based on a message bus
architecture (Tody, 1998) that connects the various components of the DHS.
One such component is the data capture agent (DCA). The DCA receives streams
of readout data from the acquisition system and builds a data object which is
shared by other components of the DHS. The image data is stored on disk as a
multiextension FITS file (MEF) (see Valdes 1997b).
The DCA is a general purpose network­based data service, using the mes­
sage bus as the network interface. It receives request messages of various types
and, through an event loop, dispatches each request to the appropriate event
handler. The request message types currently implemented in the DCA are for
control, setting or querying server parameters, readout status, data format con­
figuration, and writing header and pixel data. In addition to servicing client
requests, the DCA can broadcast messages to inform clients of the status of the
DCA during readout.
The major subsystems of the DCA at present are the message bus interface,
an event loop and event handlers for processing message requests, a distributed
shared object implementation, a keyword database system, the TCL interpreter,
and a TCL­based keyword translation module.
1 National Optical Astronomy Observatories, operated by the Association of Universities for
Research in Astronomy, Inc. (AURA) under cooperative agreement with the National Science
Foundation.
2 http://www.noao.edu/kpno/mosaic/
120

The Mosaic Data Capture Agent 121
2. DCA User Interface
The DCA may be started by issuing a command at the host level, or invoked as a
service via the message bus. This can occur during login to an observing account,
using a window manager menu, or by interactively issuing a user command. The
command invocation can include DCA server parameters which can also be set
or reset after invocation via messages.
The DCA automatically connects to the message bus when it is started.
The runtime operation of the DCA is monitored and controlled by clients via
the message bus. In the Mosaic system there are two clients; the data feed
client (DFC) and the DCA console client (DCAGUI). In general any number
of clients can access the DCA. Multiple data feed clients or GUIs can be active
simultaneously.
The DCA console client has a graphical user interface based on the IRAF
Widget Server (Tody, 1996). It can send control and parameter messages to the
DCA and act on messages broadcast by the DCA, e.g., to display the readout
status. Currently the DCAGUI executes an autodisplay command for real time
display during readout (see Valdes, 1998) and a post­processing command for
logging, archiving or other operations once the readout ends.
3. The Readout Sequence
An observation readout is driven by a sequence of DCA message requests. A
readout sequence is initiated by the DFC with a message that includes a unique
sequence number. Once a readout has started, each subsequent message asso­
ciated with that readout must be tagged with the sequence number for that
readout. Multiple readouts can be simultaneously in progress. Each readout
sequence has a separate context identified by its sequence number.
When a readout sequence is initiated the DCA creates a new image­class
distributed shared object (DSO). The DFC passes information about the size and
structure of the object (such as the number of image extensions), allowing the
DCA to configure the DSO. At this point the DFC can begin sending pixel and
header messages. At image creation time the DCA broadcasts a message so that
clients like the DCAGUI can access it if desired. Currently the DCAGUI uses
this to initiate an interim real­time display.
When the readout is completed the DCA executes a keyword translation
module (KTM), an externally supplied, interpreted TCL script which converts
detector specific information into standard header keywords. After the KTM
finishes the DCA and DSO format the output keywords, write them to the
image headers, and generation of the new image DSO is complete. The DCA
broadcasts a message when the DSO is complete which can be used to trigger
post­processing of the data.
3.1. Pixel Messages
The pixel data messages contain blocks of raw, unprocessed detector pixel data
organized into one or more streams, one for each CCD amplifier. Each stream
directs pixels to a region of an output image extension. This structure allows
the data block to simultaneously contain data for several di#erent regions and

122 Tody and Valdes
the data can be arbitrarily interleaved, encoded, flipped, or aliased. Each block
of data is processed asynchronously but the client can send a synchronization
request periodically to check the output status.
3.2. Keyword Messages
The DCA receives header information via the message bus. This information
consists of blocks of keywords organized into named, detector specific keyword
groups. The keywords are stored in keyword databases (an internal random ac­
cess data structure), one per keyword group. The set of keyword group names is
arbitrary. Some examples for the NOAO Mosaic are ICS for instrument control
system, TELESCOPE for telescope, and ACEBn for Arcon controller informa­
tion for controller n.
The Arcon controller system for the NOAO Mosaic consists of a set of
Arcon controllers each of which can readout one or more amplifiers. The current
system has four controllers each reading out two CCDs using one amplifier per
CCD. Thus there can be controller information for the whole system, for each
controller, and for each amplifier readout.
A keyword database library handles creation and maintenance of the key­
word database. Note that keyword information does not necessarily have to
come only from the DFC, the current mode of operation for the NOAO Mosaic.
Other schemes are possible.
4. Keyword Translation Module
The keyword translation module (KTM) is a TCL script called by the DCA at
the end of a readout, once all raw header information has been received. The
purpose of the KTM is to create the keywords for the global and extension
headers. The KTM is passed a list of input keyword database descriptors and
it returns a list of output keyword database descriptors, one for each header.
The DCA TCL interpreter provides special TCL commands for manipulating
(creating, accessing, searching, etc.) the keyword databases. When the KTM
finishes the DCA, via the DSO, writes the keywords in the returned output
keyword databases to the output MEF FITS file.
The KTM performs a variety of transformations on the input keywords. A
keyword can be copied verbatim if no change is desired. The keyword name or
comment may be changed without changing the value. New keywords can be
added with static information and default values may be supplied for missing
keywords. The KTM can compute new keywords and values from input key­
words. Identical keywords in each extension may be merged into a single keyword
in the global header. The KTM can detect incorrect or missing keywords and
print warnings or errors.
Two examples from the keyword translation module for the NOAO CCD
Mosaic follow.
1. The data acquisition system provides the keywords DATE­OBS and UT­
SHUT giving the UT observation date in the old FITS date format
(dd/mm/yy) and the UT of the shutter opening. The KTM converts these

The Mosaic Data Capture Agent 123
to TIME­OBS, MJD­OBS, OBSID, and the new Y2K­compliant FITS date
format.
2. The KTM determines on which telescope the Mosaic Camera is being
used and writes the WCS scale, orientation, and distortion information
previously derived from astrometry calibrations for those telescopes. The
coordinate reference point keywords, CRVAL1 and CRVAL2, are computed
from the telescope right ascension and declination keywords.
5. Distributed Shared Objects and Distributed Shared Images
An important class of message bus component is the distributed shared object
(DSO). DSOs allow data objects to be concurrently accessed by multiple clients.
The DSO provides methods for accessing and manipulating the data object and
locking facilities to ensure data consistency. DSOs are distributed, meaning
that clients can be on any host or processor connected to the message bus. In
the case of the Mosaic DHS, the principal DSO is the distributed shared image
which is used for data capture, to drive the real­time display, and for quick
look interaction from within IRAF. The distributed shared image uses shared
memory for e#cient concurrent access to the pixel data, and messaging to inform
clients of changes to the image.
The current version of the DCA does not implement the full distributed
shared image object. Instead it implements a mapped image file. The image
being created appears as a valid FITS multiextension file immediately after the
configuration information is received from the DFC. This allows applications
to examine the file while the readout is in progress. An example of this is the
interim display routine that loads a display server frame bu#er as the pixel data
is recorded, giving a simple real­time display capability.
6. Current Status and Future Developments
A version of the Data Capture Agent is in production use at two telescopes on
Kitt Peak with the NOAO CCD Mosaic Camera. The flexibility of the message
bus architecture is used to provide two modes of operation. The standard mode
uses two machines connected by fast Ethernet. One machine supports the ob­
serving user interface and the interface to the controller system. A DFC running
on this host writes to the message bus and the DCA runs on another machine
(with a di#erent OS) where the exposures are written to disk and the quick look
interaction and any post­processing are done. The second mode is a fallback
in case the second computer fails. The DCA can simply be run on the same
machine as the UI and controller interface.
Future developments will complete the distributed shared object and con­
sole client and add other clients such as the real­time display (Tody, 1997) and
data reduction agent (Valdes 1997a).
The design of the DCA (and the whole DHS based on the message bus
architecture) is open, flexible, and e#cient. It can be used with many data
acquisition systems at a variety of observatories. All that is needed is to write a
data feed client to connect to the message bus and a keyword translation module

124 Tody and Valdes
appropriate for the data. Currently the lower level messaging system is based
on the Parallel Virtual Machine (PVM) library but this can be replaced in the
future with other messaging systems such as CORBA.
References
Tody, D. 1996, in ASP Conf. Ser., Vol. 101, Astronomical Data Analysis Software
and Systems V, ed. George H. Jacoby & Jeannette Barnes (San Francisco:
ASP), 89
Tody, D. 1997, in ASP Conf. Ser., Vol. 125, Astronomical Data Analysis Software
and Systems VI, ed. Gareth Hunt & H. E. Payne (San Francisco: ASP),
451
Tody, D. 1998, this volume
Valdes, F. 1997a, in ASP Conf. Ser., Vol. 125, Astronomical Data Analysis
Software and Systems VI, ed. Gareth Hunt & H. E. Payne (San Francisco:
ASP), 455
Valdes, F. 1997b, in ASP Conf. Ser., Vol. 125, Astronomical Data Analysis
Software and Systems VI, ed. Gareth Hunt & H. E. Payne (San Francisco:
ASP), 459
Valdes, F. 1998, this volume