Integration of quality of service in avionics architecture

HAL Id: hal-02270445

https://hal.archives-ouvertes.fr/hal-02270445

Submitted on 25 Aug 2019

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Integration of quality of service in avionics architecture

G Bel, P. Bieber, F. Boniol, D Dalla Barba, G. Durrieu, B Sidoruk

To cite this version:

G Bel, P. Bieber, F. Boniol, D Dalla Barba, G. Durrieu, et al.. Integration of quality of service in

avionics architecture. Conference ERTS’06, Jan 2006, Toulouse, France. �hal-02270445�

Integration of quality of service in avionics architecture Abstract

G. Bel

, P. Bieber

, F. Boniol

, D. Dalla Barba

, G. Durrieu

, B. Sidoruk

1: ONERA-CERT, 2 av. E. Belin, BP 4025, 31055 Toulouse France 2: ENSEEIHT, 2 rue C. Camichel, 31071 Toulouse France 3: Airbus France, 316 route de Bayonne, 31053 Toulouse France

1 From federated avionics architecture to IMA

Traditionally, avionics systems have followed a federated approach - separate software functions al- located to dedicated (often physically disjoint) com- puting ”black-boxes”. As the transmitted data are involved in process control, the communication ar- chitecture between avionics functions is based on three key principles: (1) one source, one line, sev- eral receivers, (2) the transmitter does not need to know who receives data, and (3) no time synchroni- sation between transmitter and receiver. The tradi- tional aeronautical answer to these requirements was the ARINC 429 Digital Information Transfer Sys- tem: each line has only one source and is connected to every equipment that need the data transmitted by the source, and each data is individually identified (by a label) and sent. The main advantage to such architectures is determinism. However, they often lead to inefficient resource utilisation. In recent years there has been a considerable amount of effort under- taken by ARINC to define standards for Integrated Modular Avionics (IMA). IMA proposes the integra- tion of avionics systems to common resources (com- puters as well as communication lines and switches) (see figure 1).

2 The AFDX solution for communication resource sharing

The main benefits of IMA architecture are to re- duce the weight of the aircraft as less pieces of equip- ment and wiring are needed, and to reduce main- tenance cost as well as less types of spare equip- ment could be used to replace failed components.

However IMA architecture, and more specifically re-

source sharing, introduces new difficulties in design- ing and integrating avionics systems. Communica- tion resource sharing can introduce non-determinism (loss of messages, unbounded latency). The Airbus response to this problem is the new AFDX (Avionics Full Duplex Switched Ethernet) communication ar- chitecture. This architecture is based on IEEE 802.3 Ethernet and TCP/IP general principles. However it uses specific concept and features to provide secure data transfer with real time constraints, to provide the guarantee of services and determinism required for civil aerospace requirements. It is built with switches (SW), i.e., network devices in charge of data forwarding, and End System (ES), i.e., network de- vices in charge of data transmission/reception. Band- width control is achieved with advanced queue man- agement and multiple bandwidth use strategies. For that purposes, AFDX introduces the notions of BAGs (Bandwidth Allocation Gap) and of maximum frame size to allocate bandwidth as in ATM CBR (Constant Bit Rate), and UBR (User Bit Rate). This guarantee applies to the key notions of AFDX, the virtual links.

However, as far as we can not state on the ac- tual source traffic, the worst case performance of the AFDX network is systematically majored. This leads to network configurations that under-use the true Eth- ernet capability.

3 Towards resource reservation protocols and integration of Quality of Service for avionics architecture

Ethernet uses a Carrier Sense Multiple Access

with Collision Detection (CSMA/CD) arbitration

mechanism, according to which a network controller

having a message to be transmitted must wait for the

bus to become idle. This protocol does not provide,

1

ES

(Core CPM Processing

Module) (Core CPM

Processing Module)

SW SW

SW SW

SW SW

SW SW SW SW

ES ES ES

ES ES ES ES

Function A

Function B Virtual Link from A to B

(Core CPM Processing

Module)

(Core CPM Processing

Module)

(Core CPM Processing

Module)

(Core CPM Processing

Module) (Core CPM

Processing Module) (Core CPM

Processing Module)

ES

(Core CPM Processing

Module) (Core CPM

Processing Module)

SW SW

SW SW

SW SW

SW SW SW SW

ES ES ES

ES ES ES ES

Function A

Function B Virtual Link from A to B

(Core CPM Processing

Module)

(Core CPM Processing

Module)

(Core CPM Processing

Module)

(Core CPM Processing

Module) (Core CPM

Processing Module) (Core CPM

Processing Module)

Figure 1. AFDX architecture

by itself, predictable message transmission times. In order to overcome this lack of predictability, several techniques have been proposed in the last years. Ex- amples of such approaches are Master/Slave, Token- passing, Timed Token, TDMA. All these techniques operate at layer 2 (data link) by adding transmission control layer above the original MAC layer. They provide different degrees of efficiency, flexibility and timeliness guarantees. AFDX belongs to this class of techniques.

In the particular case of the Internet, two other ap- proaches have been proposed: a resource reservation approach (for example IntServ and RSVP and most of QoS general architectures), and priority based ser- vices (for examples DiffServ).

IntServ and Resource reservation protocols such as RSVP [Zha97] provide support for end-to-end resource reservation for specific sessions and de- fine a concept based on per flow reservations (sig-

nalling) and admission control to be present end-to- end. In RSVP the sessions are identified with the IP address and the ports of connects. Reservation is supported for specific sessions or clusters of ses- sions (in multicast communication). The reserva- tion protocol includes PATH and RESV messages that make the reservation of resources (especially bandwidth reservation) in all nodes (routers, switches and computers) included in the flow path. Several routers and switches support these protocols based on configurable traffic control schemas. The ba- sic problems associated to RSVP are the scalabil- ity (RSVP requires complex queuing algorithms and packet scheduling) and the wasted bandwidth neces- sary to handle the management of the whole network.

A second approach has been proposed to specifi-

cally address the problem of handling QoS require-

ments and differentiated service guarantees. Diff-

Serv is an example of such an architecture supporting

2

real-time as well as the current non-real-time service of IP and takes another approach assuming no ex- plicit reservation mechanism in the interior network elements. It defines an approach for the classification of network traffic and services and interfaces for their support. DiffServ is a layer 3 traffic-handling mech- anism including new services such as SLA (service level agreement), which specify the amount of cus- tomer traffic that can be accommodated at each ser- vice level.

It is well admitted that this last technique, Diff- Serv, combined with controlled traffic generation, provides correct result in the Internet Context for voice or multimedia traffic. The aim of this article is to evaluate these techniques (DiffServ and controlled traffic generation) in the context of avionics systems, and more particularly for embedded switched Ether- net networks.

In order to validate the Diffserv concept a lab- oratory mock-up based upon 2 COTS Gigabit Eth- ernet switches and several PC stations instrumented with traffic generation and capture capability means addressing and supporting QoS concepts and mech- anisms, was built. A first set of experiments had been conducted in order to characterize the lim- its of performance achieved with the traffic gener- ated throughput and the captured traffic. After that, throughput evaluations have been made for priori- tized traffic according to the DiffServ concept, and handling of several bandwidth reservations for prior- itized traffic has been characterized.

The results obtained show good conformance with traffic classification, traffic prioritization and band- width reservation.

3