Since 2007, CRAN has maintained a lasting partnership with CNES (Centre National d’Études Spatiales). The objective is to study the replacement of the old military communication bus dating from the 70s on board launchers, in favor of a COTS (Component On The Shelves) type network based on standards (Ethernet), more efficient in terms of throughput, reliability, cost, …
This communication system must not only match its predecessor in terms of real-time for control, but also support new applications, such as telemetry or video traffic routing. It must of course also meet the specific constraints of the space domain: low weight, robustness to stage separation, rapid reconfiguration, ability to observe all traffic in order to be able to analyze it afterwards by replaying flights.
The first studies led to the selection of a switched Ethernet standard network (figure 37). It is a real technological breakthrough that leads to the use of a non-deterministic system to support a real-time application, and opens the opportunity to break with centralized control in favor of a modular and more responsive distributed control.
By applying mathematical techniques of the "network calculus" type, we have shown that these constraints could be met by replaying past flight scenarios on this type of architecture. We have also developed fast reconfiguration techniques, compatible with the requirements of control in the event of stage separation or failure of a link or switch. We then implemented service classification mechanisms to consider supporting additional telemetry or video traffic, preserving the guarantee of proper routing of control traffic.
Finally, we have imagined a non-invasive device for comprehensive observation of all traffic allowing the recording of all communications during a flight. This last phase raised significant clock synchronization issues.
Thus, we have reached a Technological Readiness Level TRL3. It requires analytical studies and laboratory experiments validating forecasts on the separate elements of the technology, and includes components that are not yet integrated or representative.
We decided to move to TRL4 level for which the basic technological components (hardware and software) are integrated to establish that all parts work together. A full-scale communication system was therefore implemented, with components that are not "space-hardened", but implementing all the "bricks" individually validated at TRL3 level (figure 38). The demonstrator is composed of:
— 8 Cisco IE 3000 switches (implementing the IEEE1588/PTP synchronization protocol);
— 1 Cisco 2950 series switch is used to emulate the ground/air link;
— 93 micro-PCs (Raspberry PI type) used for traffic generation (86 on board and 7 on the ground);
— 8 PCs containing two Ethernet cards (including a Meinberg PTP270-PEX PTP-IEEE1588 card) for observability;
— 124 category 6 Ethernet cables (of variable length) for a total length of 192.5 m.
With this device, we can go beyond validating our proposals by simulation, and highlight the shortcomings of network equipment currently available on the market.
Ethernet standardization has evolved, with several amendments (IEEE 802.1AB, AS, ASbt, AQ, CB, Qca, Qaz, Qat, Qav, Qbv, Qch, Qci, Qbu, Qcc, Qbp, Qbz), grouped since 2014 under the term Time-Sensitive Networking (TSN). TSN aims to provide Ethernet with intrinsic real-time capabilities, allowing control with a latency of a few nanoseconds, guaranteeing delivery and jitter through the Ethernet standard. Also, the demonstrator was enriched by the purchase of 5 Cisco "Industrial Ethernet 4000" switches integrating this evolution.
The network is classically observable only locally via "mirror" ports on the switches where products like TAPs (Terminal Access Points) allow capturing traffic directly (and passively and transparently) on the links. The challenge for control and observability is to reduce measurement latencies, detection and clock offsets. Thus, 8 TAPs were acquired and installed on the demonstrator.
