Research On Scheduling And Forwarding Mechanism In Time-Sensitive Networking

With the development of the Industrial Internet,an increasing number of time-sensitive applications require deterministic transmission,i.e.,bounded and ultra-low la-tency for end-to-end communication.The standard Ethernet fails to realize deterministic transmission due to its inherent nature of the best-effort manner.Faced with this situation,Time-Sensitive Networking(TSN)is proposed by IEEE 802.1 TSN group.Composed of a series of standards on the time synchronization,the packet scheduling and forwarding mechanism,as well as the network management and configuration,TSN aims to not only guarantee the deterministic transmission requirements for time-sensitive traffic,but also provide the Quality of Service(Qo S)for time-insensitive traffic.TSN is now developing rapidly,and there are lots of technical challenges to be addressed.This dissertation focuses on improving the performance of the packet scheduling and forwarding mechanism in TSN,and achieves the following contributions:1.A light-weight,fast-response and full-coverage monitoring system TSN-Guard is proposed to detect and report the forwarding misbehaviors of time-sensitive traffic.Even though several standards have been drafted in order to try to guarantee the bounded and ultra-low latency for time-sensitive traffic,the expected performance cannot always be achieved because of the packet forwarding misbehaviors.To monitor the performance of a TSN network in real-time,this dissertation first makes a penetrating analysis on the causes and characteristics of forwarding misbehaviors,and then elaborates solutions to build an on-line tracking system,including the misbehavior identifier and data collecting process.Experimental results show that the comprehensive performance of TSN-Guard is far superior to the existing approaches.For example,the bandwidth overhead is only 11%of the event-driven approach,and the average response time is only 49% of the fixed-cycle approach.2.To improve the guard-band utilization,this dissertation proposes an algorithm family of Packet-size Aware Shaping(PAS).The IEEE 802.1Qbv standard specifies guard-band to ensure deterministic packet forwarding,but the bandwidth utilization in guard-bands is low.PAS tries to select eligible packets to utilize the guard-band.By abstracting the problem of utilizing the guard-band to a classic Precedence-Constrained Knapsack Problem(PCKP),several specific PAS algorithms are proposed,i.e.,the ideal PAS(IPAS),the greedy PAS(GPAS)and the 8)-set GPAS(8)-GPAS).Markov Chain is applied to model the problem mathematically.Analytical performance is verified by the simulations.The results indicate that using 4-GPAS in real networks can achieve satisfy-ing comprehensive performance.3.To achieve high scheduling speed while cutting down hardware consumption,an improved programmable hardware-based packet scheduling framework is proposed.The corresponding hardware Push-In-Pick-Out(PIPO)prototype is designed and demon-strated accordingly.The new framework can not only flexibly express the scheduling algorithms specified in the standards,but also support a variety of customized scheduling algorithms.The PIPO prototype consists of simple priority queues to approximate the state-of-the-art Push-In-Extract-Out(PIEO)prototype.Experimental results on FPGA show that the PIPO prototype can save considerable on-chip hardware resources towards achieving high-speed throughput and good scalability.For example,with 2,048 input queues,the PIPO prototype achieves a throughput of 70 Mpps,which is 1.64 x higher than the PIEO prototype,but using only 14.7% Look-Up Tables(LUTs)and 40.5% Block RAMs of the latter.4.A name prefix caching approach,Plus-Bitmap Caching(PBC),is proposed.Nowadays,Open Platform Communication Unified Architecture(OPC UA)over TSN is a promising direction in Industry Internet.Current OPC UA over TSN uses the TCP/IP protocol stack,which unfortunately cannot support the publish/subscribe model in OPC UA efficiently as it is not designed for multicast.To solve this problem,the name-based routing in Named Data Networking(NDN)is a promising solution.To explore the advan-tages of combining NDN and TSN,this dissertation proposes PBC to accelerate the name lookup process.Innovatively,PBC uses a bitmap structure to solve the cache hiding and consistency problems.A complete software-based name-lookup system is constructed and theoretical analysis as well as extensive experiments are conducted.Experimental results show that PBC outperforms other approaches on cache hit ratio(going beyond up to 16%).Further,the PBC-based system can effectively improve the lookup and update speed.