Experimental Evaluation of the Layered Flow-Based Autonomous TSCH Scheduler

Abstract

The Industrial Internet of Things (IIoT) requires wireless connectivity that meets the strict industrial requirements on metrics such as reliability and latency. Promising approaches include Time Slotted Channel Hopping (TSCH) media access, where nodes operate according to a schedule. Autonomously built schedules typically rely on shared resources, where reliability and latency may suffer depending on traffic scenarios and topologies. We have earlier proposed the Layered scheduler, which belongs to a new category of autonomous schedulers: Flow-based scheduling. Layered allocates resources to traffic flows, and as opposed to typical autonomous schedulers, dedicated resources are guaranteed to be scheduled at every hop from source to destination in a convergecast scenario. In addition, Layered minimizes the number of channels required through the novel employment of autonomous spatial reuse. We extend earlier theoretical analysis and simulations by evaluating Layered using the FIT IoT-LAB testbed and compare it to Orchestra and 6TiSCH Minimal scheduler. The experiments demonstrate the feasibility of spatial reuse and that Layered retains performance independent of network topology and traffic intensity -a desirable feature in industrial scenarios. The performance comes at the expense of energy consumption, which in the worst case is 75 % higher compared to Orchestra and Minimal. We also present lessons learned, such as the crucial relationship between TSCH and RPL convergence, the benefits of black-listing on performance, and how co-located TSCH networks could be divided by channel offsets as opposed to physical channels. Lastly, we discuss flow-based scheduling in general, its properties, and future research areas.

Autonomous Flow-Based TSCH Scheduling for Heterogeneous Traffic Patterns: Challenges, Design, Simulation, and Testbed Evaluation

“…in the case of control applications employing a sensor-to-actuator pattern. We modified our 18 See Section VII.B in [3] autonomous flow-based Layered scheduler to support heterogeneous traffic patterns. The modifications included a novel mechanism to learn which cells to schedule autonomously.…”

“…We proposed the autonomous Layered scheduler in [15] and conducted testbed evaluations in [3]. Layered operates in a flow-based manner by allocating cells to traffic flows that are assumed destined for the RPL root.…”

“…To install the correct cells in its schedule, a node needs to know 1) Its own depth in the routing tree, which may be 2 Number of hops from the root in the RPL tree learned from RPL, and 2) The IDs of flows passing through the node. We assume each node in our sub-tree generates one flow each, and we let every node's unique ID serve as flow identifier 3 . The last byte of each node's IP address reflects the node ID.…”

“…The duty cycles of each scenario are seen in Figure 11. Although the longer timeout risk stale allocations incurring 16 We identified the best channels as 18 -21 in earlier work [3] 17 Packet Reception Rate measures the quality of a link by taking the number of received acknowledgments to the number of transmission attempts.…”

“…Autonomous schedulers can be categorized based on whether they assign cells to nodes, links, or traffic flows. In : [3], we argued that flow-based schedulers benefit industrial applications as they can guarantee resource allocation end-toend for each traffic flow. In addition, they provide flexibility such as scheduling multiple flows from one node, and may open for flow concepts as in IETF's Deterministic Networking (DetNet) [4].…”