Technischer Bericht TR-2016-03

Nayak, Naresh Ganesh; Dürr, Frank; Rothermel, Kurt: Time-sensitive Software-defined Networks for Real-time Applications.
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik, Technischer Bericht Informatik Nr. 2016/03.
24 Seiten, englisch.
CR-Klassif.C.2.5 (Local and Wide-Area Networks)

Cyber-physical systems (CPS), like the ones used in industrial automation systems, are highly time-sensitive applications demanding zero packet losses along with stringent real-time guarantees like bounded latency and jitter from the underlying network for communication. With the proliferation of IEEE 802.3 and IP networks, there is a desire to use these networks instead of the currently used field-buses for time-sensitive applications. However, these networking technologies, which originally were designed to provide best effort communication services, lack mechanisms for providing real-time guarantees. In this paper, we present Time-Sensitive Software-Defined Networks (TSSDN), which provide real-time guarantees for the time-triggered traffic in time-sensitive systems while also transporting non-time-sensitive traffic. TSSDN provides these guarantees by bounding the non-deterministic queuing delays for time-sensitive traffic. To this end, TSSDN exploits the logical centralization paradigm of software-defined networking to compute a transmission schedule for time-sensitive traffic initiated by the end systems based on a global view. In particular, we present various Integer Linear Program (ILP) formulations for computing high-quality transmission schedules. Moreover, we show that end systems can comply with a given schedule with high precision using user-space packet processing frameworks. Our evaluations show that TSSDN has deterministic end-to-end delays (<=14 us on our benchmark topology) with low and bounded jitter (<= 7 us) for packets of size 1500 bytes transmitted with a frequency of 10 kHz.

Volltext und
andere Links
PDF (1086179 Bytes)
Abteilung(en)Universität Stuttgart, Institut für Parallele und Verteilte Systeme, Verteilte Systeme
Eingabedatum3. Mai 2016
   Publ. Abteilung   Publ. Institut   Publ. Informatik