Time-aware Traffic Shaper using Time-based Packet Scheduling on Intel I210

DOI®: doi.org/10.21276/ijre.2018.5.9.1

CITATION: Alhady, S., Othman, W., Wahab, A., Wahab, A., & Wong, P. (2018). Time-aware Traffic Shaper using Time-based Packet Scheduling on Intel I210. International Journal Of Research And Engineering, 5(9), 494-499. doi:10.21276/ijre.2018.5.9.1

Author(s): 1Syed Sahal Nazli Alhady, 2Wan Amir Fuad Wajdi Othman, 3Aeizaal Azman Abd Wahab, 4Aeizaal Azman Abd Wahab, 5Por Yin Wong

Affiliation(s): 1,2,3,4,5School of Electrical & Electronic Engineering, Universiti Sains Malaysia


By 2015, the Institute of Electrical and Electronics Engineers (IEEE) time-sensitive networking (TSN) task group has released several TSN standards. Amongst them is 802.1Qbv, also known as a time-aware shaper, aiming to provide performance assurances of latency and delivery variation to enable applications in a TSN network. While there are several products and evaluation kits that employ 802.1Qbv in the market now, it is still not widely adopted yet due to the maturity of the standard. Hardware-enabled 802.1Qbv use hardware queues and timers to achieve accurate transmission of packets in the switch and bridge. This research aims to investigate the feasibility of using an existing end-station Ethernet controller, Intel I210, and its launch time control feature (commonly known as time-based packet scheduling) to shape traffic compatible to 802.1Qbv-enabled network bridges. A software solution is developed by implementing a software configurable gate-control list and employing open-source Linux RFC patches for per-packet transmit time specification. By configuring the kernel and mapping kernel-layer traffic classes to the hardware queues, packets can be transmitted out at precise times while attaching 802.1Q VLAN tags, required by bridges to identify packets. Through analysis, it is found that this solution will require an additional 30 ?s transmit offset to be used effectively. That is 55% more time is needed to transmit a packet in a back-to-back connection and 17.6% on a 3-switch network to improve period peak jitter performance to just 8.9  ?s compared to 1 ms on solutions that send packets out periodically using software sleep functions.
Public Knowledge Project [ INDEXED LINK at index.pkp.sfu.ca ]

Zenodo Research Platform [ INDEXED LINK at zenodo.org ]

Indexed by other automated sites of Open Journal System under Public Knowledge Project http://pkp.sfu.ca/ojs

  • 1.4K