The following sections outline our intended equipment setup for this acquisition task.
We suggest using a standard PC equipped with a custom interface card running Linux.
The main task in this PC is to capture and de-frame HRDL data and write all of it immediately onto hard disks. Statistics of detected errors (please see section errors ) are updated and at least displayed, and possibly made available to the GSE via log files. A standard FTP server makes data and log files available to GSE on-line analysis computers. Alternatively, it is possible to expressly send new files to a FTP server residing on a GSE computer.
We recommend using a dedicated network interface in both PDS and GSE computers for HRDL data transfer traffic. Choosing twisted pair Ethernet as this transfer media allows us to use 10/100Mbits/s selectable network cards.
The unique file cache memory management of Linux means that by increasing the amount of main memory in PDS computer it is possible to avoid practically all disk accesses required to transfer just-captured data via FTP to GSE. Under Linux all main memory which is not in active use by programs will be used to buffer recently-written and/or recently read disk data.
DAT tapes can be used to archive old data if 46GBytes of hard disk space proves to be too expensive to store all of the data to be received in 100h shuttle mission. The slowest DAT tape drives offer a throughput of more than 150kBytes/s (typically 180kBytes/s) which is just barely enough to store 128kBytes/s continuously. Modern DDS2 drives usually offer transfer rates in excess of 500kBytes/s which should be sufficient to write tapes only during off-halves of the duty cycle.
We would expect that one subdirectory full of data files made during one cycle of Acquire-Loss of Signal (AOS/LOS) would be a suitable candidate for partitioning the data into smaller units. It would make sense to use three subdirectories under each AOS/LOS session subdirectory, perhaps as follows:
rec:The directory for receiving new data. Once a
file is ready for FTP, it will be moved to the ``ftp''
subdirectory. Creating a new file with 30--60 second intervals would
result in files being about 8--16MBytes each and this would provide
the required almost immediate access to new data.
ftp:Files which have not yet been transferred but
which are ready for transmission. Once successfully transferred, the
GSE computer can remotely move the files to the ``arc''
subdirectory.
arc:Files which have been consumed by the GSE on-line analysis system and which can thus be archived onto DAT tape. Once successfully archived, these files can be removed to make room for new files in the on-line hard disk buffer.
Moving files from one directory to another is a lightweight operation under Linux, and at the same time it is atomic, i.e.~it can be used to synchronize data capture, FTP, and DAT archiving processes running in parallel. One AOS/LOS cycle will result in approximately 500MBytes of data, in about 50 files of about 10MBytes each in one subdirectory which seems like a reasonable file setup.
The strategy here is to capture all bits as they come off the HRDL and write them immediately onto a buffer hard disk. We would expect that one on-half of the duty cycle would be a suitable unit for buffering, i.e.~a unit to write on hard disk and later onto DAT tape.
We can use a similar interface card as in the primary data storage computer, but in this application all sync word detection circuitry will be bypassed and only the simple shift registers and bus interfaces will be used.
A secondary level of backup is provided by NASA tapes made of shuttle downlink data. These tapes are rumored to be difficult to decode, and thus these are intended only for protection against catastrophic failures in both primary storages at the same time.
We consider the following to be two major reasons promoting this setup as a practical way to acquire HRDL data.
Next Chapter, Previous Chapter
Table of contents of this chapter, General table of contents
Top of the document, Beginning of this Chapter