IP In Space
IP In Space
BY WILLIAM JACKSON
| GCN STAFFNASA wants Earth protocols to work over longer distances
As government and commercial networks reach out into space, NASA is looking for protocols that can transmit data to satellites, the International Space Station and beyond.
Engineers at the agency's Glenn Research Center in Cleveland are tinkering with TCP/IP, the ubiquitous protocol suite, to see what parts need fixes for space and wireless communications.
'The tools are there and should work,' senior research engineer William Ivancic said. 'What we're doing at Glenn is making sure it does work.'
So far, most of it does.
'For near-Earth stuff, you can use just about everything out of the box,' he said. 'There is no reason you can't control an experiment on the space station from the ground over IP.'
But TCP/IP is not ideal for NASA's needs, and the engineers are working on a few improvements they hope will become standard. They want the changes incorporated in the off-the-shelf protocol stack, not modified for their use, 'because as soon as you modify, you lose all the commercial support,' Ivancic said.
NASA has had plenty of experience working with proprietary protocols, which cost a great deal to maintain.
Space communication usually has been via point-to-point radio links tailored to each mission.
'Mostly it's one-off stuff,' Ivancic said. Links have been expensive to maintain and unsuited for exchanging information among multiple nodes.
'We're trying to characterize the protocols'figure out where they can be used and where they shouldn't be,' he said.Phone home
Astronauts tested voice over IP telephones for personal conversations during the last shuttle mission [GCN, March 5, Page 1
]. The protocols were designed for relatively short links, however, and they rely on feedback to optimize performance.
TCP waits for an acknowledgment after sending a packet. When initial packets start to arrive, it picks up the pace of transmission. If packets do not come promptly, TCP assumes the reason is network congestion and slows down the transmission rate. The resulting latency is too great for long space hops.
TCP/IP 'won't work in deep space because of latency,' Ivancic said.
NASA Glenn scientists are testing the protocols together with engineers from the Federal Network Systems unit of Verizon Communications Inc. They send data to satellites in geostationary orbit 22,000 miles above Earth and model the expected performance for longer links to the moon, Mars and other nearby planets.
'Communicating with a satellite, there is a half-second round-trip delay in the control loop,' said Mark Allman, a network engineer in Verizon's federal group. 'If you extrapolate to a lunar orbital satellite, the handshake duration becomes more severe. The control loop is about four minutes.'
The point at which latency gets too great 'depends on what you are trying to do,' Ivancic said.
NASA's William Ivancic says the space agency is avoiding custom apps because a standard, unmodified protocol for space communications will ensure commercial support.
The TCP/IP suite is flexible enough to overcome some difficulties. The User Datagram Protocol, a connectionless protocol running under IP, can solve some of the handshake problems but does very little error correction. NASA is looking for intelligent applications that will not need error correction.
If the protocols have to be tweaked, NASA wants changes that will be acceptable as standards.
'You can always optimize a protocol,' Ivancic said. 'The problem is, it's only optimized for your situation, and you lose support.' Internet experts look for the lowest common denominator, asking if it's good enough for a standard, he said.The right stuff
Glenn researchers have come up with several ideas for standards that Verizon technology officer Bruce E. Fleming said 'look very promising.'
'We're close on some things,' Ivancic agreed.
Some of the possible standards include TCP Pacing, by which the sending end assesses available bandwidth to time packet release, and Packet Pair Probing, which helps both sender and receiver intelligently determine network conditions to improve bandwidth use. Also, Explicit Transport Error Notification lets a sender know when a packet is lost in error rather than from congestion. Error notification would prevent slowdowns because a TCP sender assumes the network is busy.
'We're at the beginning of the research, and it looks like you can get a lot of improvement,' Ivancic said.