A real-time multicast routing scheme for multi-hop switched fieldbuses

Abstract

Abstract-The rapid scaling up of Networked Control Systems (NCS) is forcing traditional single-hop shared medium industrial fieldbuses (a.k.a. fieldbuses) to evolve toward multi-hop switched fieldbuses. Such evolution faces many challenges. The first is the re-design of switch architecture. To meet the real-time nature of NCS traffic, and to lay a smooth evolution path for switch manufacturers, it is widely agreed that a (if not the) promising switch architecture is an input queueing crossbar architecture running TDMA scheduling. The second challenge is real-time multicast. NCS applications usually involve complex distributed multipleinput-multiple-output interactions, which by their nature necessitate real-time multicast. In shared medium fieldbuses, real-time multicast is straightforward as data sent to the medium is heard by all nodes. On multi-hop switched fieldbuses, however, real-time multicast becomes non-trivial. In this paper, we prove real-time multicast on multi-hop switched fieldbuses is NP-Hard. What is more, real-time multicast on multi-hop switched fieldbuses is fundamentally different from Internet multicast, due to realtime requirement and the homogeneous input queueing crossbar switch architecture. Particularly, switch external links' capacities are no longer mutually independent. Such drastic change of assumptions warrants developing new routing algorithms, and a heuristic algorithm is hereby proposed.

Curbing Aggregate Member Flow Burstiness to Bound End-to-End Delay in Networks of TDMA Crossbar Real-Time Switches

“…Later in this section, we will show due to insufficient flow isolation, this approach makes the analysis of each flow's [8], note C 2 and C 3 may be fetched/outputted in different cell-times, but this does not matter). schedulability (i.e., how much resources must be allocated to guarantee the existence of an E2E delay bound?)…”

“…(8). Proof: Please see Appendix C. Note the preconditions in Theorem 2 and 3 define the constraint on all clock drifts between switches: τ max −τ min < τ min /M .…”

“…A (if not "the") widely recognized/adopted real-time switch architecture for multi-hop real-time networks is the TimeDivision-Multiple-Access (TDMA) crossbar real-time switch architecture (simplified as "TCRT switch" architecture in the following) [7] [8][1] [9][10] [11]. This architecture is particularly important for mainstream switch manufacturers because it complies with (and even simplifies) a mainstream Internet switch architecture [1].…”

“…So far, we are assuming all flows are unicast. The extension to support multicast is simple [8], as shown in Fig. 3.…”

Analysis of TDMA crossbar real-time switch design for AFDX networks

“…[14]): in this example, the switch has N = 4 inputs and outputs, frame size is M = 5 slots (note in reality, M is in the order of 10 3 ∼ 10 5 ); each row of the "schedule matrix" is a conflict free schedule for its corresponding output, which means at any time slot (i.e., any column of the "schedule matrix"), no two outputs contend for the same input (for different input queues).…”