Mixed criticality management into real-time and embedded network architectures : application to switched ethernet networks

Abstract : MC (Mixed-Criticality) is an answer for industrial systems requiring different network infrastructures to manage informations of different criticality levels inside the same system. Our purpose in this work is to find solutions to integrate gls{MC} inside highly constrained industrial domains in order to mix flows of various criticality levels inside the same infrastructure. This integration induces isolation constraints : the impact of non-critical traffic on critical traffic must be characterized and bounded. This a condition to respect timing constraints. To analyze transmission delays and focus on the determinism of transmissions, we use an end-to-end delay computation method called the trajectory approach. In our work, we use a corrected version of the trajectory approach taking into account the serialization of messages.To assure the respect of timing constraints in mixed critical networks, we first present a theoretical model of gls{MC} representation. This model is issued from gls{MC} tasks scheduling on processors. This model proposes a flow modelization which considers that each flow can be of one (low critical flows) or several criticality levels.To integrate gls{MC} inside gls{RT} networks, we propose two network protocols. The first is the centralized protocol. It is structured around the definition of a central node in the network, which is responsible for synchronizing the criticality level switch of each node through a reliable multicast protocol in charge of switching the network criticality level. This centralized protocol proposes solutions to detect the needs to change the criticality levels of all nodes and to transmit this information to the central node. The second protocol is based on a distributed approach. It proposes a local gls{MC} management on each node of a network. Each node individually manages its own internal criticality level. This protocol offers solutions to preserve when possible non-critical network flows even while transmitting critical flows in the network through weak isolation.In order to propose an implementation of these protocols inside Ethernet, we describe how to use Ethernet 802.1Q header tag to specify the criticality level of a message directly inside the frame. With this solution, each flow in the network is tagged with its criticality level and this information can be analyzed by the nodes of the network to transmit the messages from the flow or not. Additionnally, for the centralized approach, we propose a solution integrating gls{MC} configuration messages into gls{PTP} clock-synchronization messages to manage criticality configuration information in a network.In this work, we designed a simulation tool denoted as gls{ARTEMIS} (Another Real-Time Engine for Message-Issued Simulation). This tool is dedicated to gls{RT} networks analysis and gls{MC} integration scheduling scenarios. This tool, based on open and modular development guidelines, has been used all along our work to validate the theoretical models we presented through simulation. We integrated both centralized and decentralized protocols inside gls{ARTEMIS} core. The obtained simulations results allowed us to provide information about the gls{QOS} guarantees offered by both protocols. Concerning non-critical traffic : the decentralized protocol, by permitting specific nodes to stay in non-critical nodes, assures a highest success ratio of non-critical traffic correct transmission.As a conclusion, we propose solutions to integrate gls{MC} inside both industrial and gls{COTS} Ethernet architectures. The solutions can be either adapted to clock-synchronized or non clock-synchronized protocols. Depending on the protocol, the individual configuration required by each switch can be reduced to adapt these solutions to less costly network devices
Document type :
Theses
Complete list of metadatas

Cited literature [138 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01708299
Contributor : Abes Star <>
Submitted on : Tuesday, February 13, 2018 - 3:03:07 PM
Last modification on : Wednesday, July 4, 2018 - 4:37:59 PM
Long-term archiving on : Monday, May 7, 2018 - 6:02:13 PM

File

TH2016PESC1033.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01708299, version 1

Citation

Olivier Cros. Mixed criticality management into real-time and embedded network architectures : application to switched ethernet networks. Operations Research [cs.RO]. Université Paris-Est, 2016. English. ⟨NNT : 2016PESC1033⟩. ⟨tel-01708299⟩

Share

Metrics

Record views

260

Files downloads

405