Steve Scott [1], Darrel Emerson [2], Rick Fisher [3], Mark Holdaway [2], Jill Knapp [4], Lee Mundy [5], Remo Tilanus [6], Mel Wright [7]
November 13, 1996
Keywords: computing, software, pipeline, data rate
The computing problem for the MMA is analyzed with the goal of providing the requirements and specifications that directly affect the scientific capabilities and that define the scope of the computing task. The vision that is adopted is that from the perspective of the astronomer the MMA will appear as a remotely controlled imaging instrument. It will be possible to control the MMA from basically anywhere and the output for the majority of projects will be an automatically generated image cube. Image generation is done with a multi-processor imaging pipeline.
Certain instrumental specifications that have a strong effect on the computing problem are called out. These include maximum and average data rate, minimum integration time for both interferometric and single dish observations, and the maximum setup time. The minimum integration and maximum setup times will require coordination with the hardware designers. The maximum data rate of 10MB/sec (0.9TB/day) is chosen to be consistent with the science goals presented at the Tucson workshop. Data recording at the maximum rate is not difficult, but for the pipeline to be realizable, an average data rate of 10% of the maximum is specified.
Different modes of using the MMA (e.g. synthesis imaging, single dish position switching) are enumerated as well as different ways of grouping the telescopes. Methods for realtime calibration and processing of the data are also presented. The imaging pipeline is found to be a challenging aspect of the MMA as it requires a large amount of raw computing power. Detailed benchmark results are used as the basis for estimating the pipeline computing requirements with the result that the equivalent of 40 workstation level processors (at year 2005 performance levels) will be needed.
Instrument capabilities include use as a remotely controlled classical interactive telescope, such as today's single dish millimeter telescopes, or control with a flexible queue that schedules the telescope in reaction to atmospheric conditions. An innovative "checkpoint" concept is also explored that extends the power and flexibility of interactive use. Other additions to the online and offline software are proposed that will make the MMA more powerful and easy to use.
[1] Caltech/OVRO
[2] NRAO/Tucson
[3] NRAO/GB
[4] Princeton
[5] BIMA/UMD
[6] JCMT
[7] BIMA/Berkeley
View an HTML version of MMA Memo #164.
Download a 16.4kB gzipped postscript version of MMA Memo #164 from
Last modified: 09 December, 1999
kweather@nrao.edu