Spacecraft Computers: past, present, future; and gnu/Linux in Space

1
Spacecraft Computerspast, present, futureand
gnu/Linux in Space

• Patrick H. Stakem
• Sheffield, UK
• October 2003

2
As one of my predecessors said,

• "I cannot find words to express how deeply I feel
  the honor of addressing some of the foremost
  thinkers of the present time, and some many able
  scientific men, engineers, and electricians, of
  the country greatest in scientific achievement."
• Lecture before IEE, Feb. 1892, Nikola Tesla

3
disclaimer

• The opinions expressed are those of the author,
  and do not necessarily represent NASA or QSS
  policy.

4
What is a flight computer?

• Spacecraft onboard computer
• radiation hardened
• low power wide temperature range
• embedded
• no rotating secondary memory
• Custom versus c.o.t.s.
• Tasks
• communications and data handling
• attitude and orbit control
• power and thermal management
• instrument control

5
Past

• Earliest computers were much too big and heavy to
  be placed onboard. (1950s and 1960s)
• Mainframes missile guidance computers
• Simple dedicated hardwired controllers
  (1960s-70s)
• general purpose computing power, radiation
  hardened logic gates (1970s-1980s)
• COTS hardware, carefully selected, (1990s)
• COTS/Open Source software 2000

6
NASA Using free Softwareon Earth and in Space

• Linux
• workstations
• FlightLinux
• Beowulf
• TCP/IP protocols
• IP-in-space
• Interplanetary Internet

7
(No Transcript)
8
Issues in Using Free Software

• Data security and privacy
• Verification, for Mission critical applications
• Policy - management view and understanding
• Lack of the cost metric
• if its free, its not worth anything
• Experience base

9
FlightLinux

• A New Option for Spacecraft
• Onboard Computer Operating Systems

10
FlightLinux Project

• Selected by NASA Headquarters, Office of Earth
  Science in May 2000 as a multi-year funded
  project.
• Government-Industry team
• Principal Investigator Pat Stakem, QSS Group,
  Inc.
• Partners
• Surrey Space Technology Labs (UK)
• Omni Project (NASA/Goddard Code 588)

11
FlightLinux onboard computer Linux port Assessment

• Target base architecture assessment
• RAD6000 R/6000 - PPC-603e cots
• RH32 MIPS, R3000 cots
• Mongoose-V MIPS, no MMU cots, modified
• RHPPC PPC cots
• RAD750 PPC-750 cots
• ERC32 SPARC cots
• IA-32 Pentium, 80x86 cots
• SNAP-1 StrongARM cots
• here, cots a Linux version exists.

12
Benefits

• Onboard LAN
• FireWire/SpaceWire
• 1553/1773 Master/Slave
• 10Base-T
• Onboard file system, in the bulk memory
• Onboard Java applets, via JVM
• Onboard web page serving
• IP to and on the spacecraft

13
FlightLinux Project Web Page

• http//flightlinux.gsfc.nasa.gov/
• The Flight Linux Project officially concluded
• on June 30, 2002. We continue to look for
  partners.

14
Related Research

• Posix-compliant application software
• End-to-end IP IP-to-the-spacecraft
• Omni Project - this has been demonstrated.
• CHIPS spacecraft
• Flight Java
• algorithm migration demonstrated
• Onboard networked file systems
• Beowulf - distributed processing

15
Related research-onboard algorithms

• FlatSat (Omni Project)
• 233 MHz Intel-based processor
• embedded system, PC-104 bus, Linux
• Multispectral image classification algorithm
• implemented in Java, 7 Megabyte footprint
• 70-90 data downlink reduction (demonstrated
  1/2001)
• Onboard LAN connected instrument
• TCP/IP over 10Base-T
• simulated scanning instrument, Landsat MSS-class
• Downlink
• TCP/IP over 10Base-T

16
(No Transcript)
17
Other space linux applications

• Space shuttle Experiment on STS-83 plant growth
  experiment Debian gnu/linux
• International Space Station ESA laptop
  experiments
• Numerous balloon and sounding rocket applications

18
Interplanetary Internet

• Extension of Internet off of the Earths surface
  to other planets
• to low Earth Orbit, IP, Mobile-IP works ok.
• need new approaches at planetary distances
• long-haul optical channels between planets
• protocols that are resilient to long delays
• Planetary infrastructure TDRSS and GPS
• IPN-SIG discussion group
• Vint Cerf, Adrian Hooke (JPL), et al
• http//www.ipnsig.org/

19
GroundApplicationsof free software

• Low cost workstations
• Beowulf clusters

20
Low cost workstationsusing gnu/linux

• Commodity pcs as an alternative to workstations
• Linux as an alternative to proprietary OpSys
• Linux apps as alternatives to commercial packages
• with due concern for interoperability of file
  formats
• When budget is a concern (or, non-existent)
• When application migration is important

21
Beowulf clusters

• Concept code developed at NASA/GSFC by Center
  of Excellence in Space Data and Information
  Sciences (Code 930.5). Now commercialized by
  Scyld Corp.
• Low-cost cluster of workstations working
  cooperatively to process science data at
  super-computing speeds.
• Public-domain and open source software (COTS)
• Linux Operating System basis
• MPI (Message Passing Interface) or PVM (Parallel
  Virtual Machine)
• More Information
• http//beowulf.gsfc.nasa.gov/
• http//www.scyld.com/

22
The QSS Beowulf Cluster

• 16 nodes single processor Pentium 133 MHz, 32
  meg RAM, 1Gbyte hard disk, CD-ROM, no keyboard,
  mouse or screen.
• 100 Mbps ethernet connection (private network).
• Only the master node has human interface, and a
  connection to the corporate LAN / internet.

Recycled Corporate PCs
23
Beowulf cluster performance

• Caveat There are no general purpose parallel
  machines the architecture interacts with the
  problem space.
• A good parallel machine can transform a
  compute-bound problem into an I/O bound problem.
• Choosing the right problem is important. We
  choose Embarrassingly Parallel problems to
  showcase the technique.
• Thus, we can get a 15 x speedup for 16 nodes, in
  a class of image processing problem.

24
Onboard Science Data Processing(OSDP)
Testbed

• As part of the task, we studied EOS-era science
  data processing demands.
• Identified a MODIS product (cloud cover) with
  potential to migrate onboard the spacecraft
  (FlatSat).
• Began development of a testbed for analysis of
  onboard science data processing using the MODIS
  direct broadcast (OSDP).
• More information
• http//aqua.qssmeds.com/osdp
• Technical report http//aqua.qssmeds.com/osdp/doc
  s/report.html

25
(No Transcript)
26
The PNN Image Classifier
27
The PNN Image Classifier
S. R. Chettri Probabilistic neural network
architecture for high-speed classification of
remotely sensed data, Telematics and
Informatics, vol. 10, No 4, pp. 187-198, 1993
28
The Training Data Set

• Ground truth provided by the USGS

29
(No Transcript)
30
Whats next?

• Flight Beowulf
• cluster computing between members of a
  constellation of spacecraft
• enabled by tcp/ip in space
• NanoSats, with swarm intelligence
• social insect behavior model
• Reconfigurable computers
• based on rad-hard FPGAs

31
In conclusion

• The future of Free Software in space looks good.
• It can be accepted in real-world applications.
• It requires a new paradigm and new policies.
• Its a lot more fun.
• Its use is not limited to the surface of one
  planet.