Using the Internet Protocol Suite for Deep Space

Space communications are challenging

Compared to Earth Internet, space communication and networking is very challenging as it involves very long delays like 10 to 40 minutes round-trip time from Earth to Mars and interruptions of communications, from minutes to hours to days, because of orbital dynamics. For example, a Mars orbiter typically will be able to communicate with a surface rover over 30 minutes every 6 hours. Every 26 months, the Sun is in the middle between Earth and Mars and therefore cut any direct communication for around 2 weeks.

Using the Internet Protocol Suite for Deep Space

The Internet Protocol suite can not be used as is. However, we were able to profile and adapt it to make it work in deep space, enabling a seamless network from Earth to Moon to Mars to deep space. The reuse of these current technologies vastly lower the costs for space missions.

Key Adaptations to the Internet Protocol Suite for Deep Space

There are 3 key adaptations to be made to the Internet Protocol(IP) suite for deep space use.

Storing Packets

First, on the Internet, when a destination becomes unreachable, such as when a link to the destination is down and there is no alternate route, the router holding the packet will drop it and will return an error message to the source using ICMP. In deep space, given that it is normal to have link interruptions such as when an orbiter is leaving the line of sight, the router should instead store the packet until the link becomes up again or an alternate route is possible. Viagénie has demonstrated a prototype for this function.

A Transport profiled for long delays

Second, on the Internet, the transport protocols used such as TCP or more recently QUIC assume that the two peers are relatively always connected and that the time for transmission is low, usually way below 1 second. Those transport protocols then are exchanging multiple packets to estimate the round-trip time and to decrease the likelihood of congestion. In space, with delays of minutes to hours to days, the latter because of link interruptions, such behaviour does not work. TCP is more difficult to adapt as it has multiple packets just for the connection establishment and is implemented in the kernel. However, the more recent QUIC now counting for a significant portion of Internet trafic is more adaptable to deep space. Viagénie has demonstrated that by setting specific QUIC parameters, HTTP transactions are possible over multiple days delays.

Applications have to be implemented for long delays

An application using HTTP typically assumes that the response will come back in matter of milliseconds. As such, HTTP clients have timeouts in terms of seconds. If an application is properly implemented by doing all its network transactions in a true and complete asynchronous way and that it sets timeouts large enough to support the deep space RTT, then the application, using either HTTP, QUIC or UDP, would just work fine. Viagénie has demonstrated that by setting proper timeout values of HTTP clients and servers and by having applications using asynchronous network transactions, one can use deep space for IP-based applications.

Testing is a Key Component

Testing is very important in any software and hardware development. But in space, it is paramount to decrease the possibilities of any issue that can have a big impact on a mission. Viagénie has created a large testbed enabling the tests of the Internet Protocol suite, from IP to transport (QUIC and UDP) to HTTP to applications to network management, time distribution and media. An important part of the work has been to automate this testbed enabling reproducible tests and results that can be shared to the technical community.