Page 44 - ITUJournal Future and evolving technologies Volume 2 (2021), Issue 1

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 1

0.45 ST packet delays. A potential workaround to explore in
Reconfiguration - ST-Max τ = 2
0.4 Reconfiguration - ST-Max τ = 3 future research is to service all the ST streams irst, and
Reconfiguration - ST-Max τ = 4
Reconfiguration - ST-Max τ = 5
0.35 No Reconfiguration - ST-Max τ = 2 then service CDT frames before servicing the BE traf ic,
Max Packet Delay (ms) 0.25 pending on the ST load.
No Reconfiguration - ST-Max τ = 3
though this might lead to additional signaling delays de‑
0.3
No Reconfiguration - ST-Max τ = 4
No Reconfiguration - ST-Max τ = 5
0.2
6.
CONCLUSIONS AND FUTURE WORK
0.15
0.1
0.05 The IEEE 802.1Qcc framework and the 802.1Qbv traf‑
ic shaper enable the implementation of a deterministic
0 forwarding plane that provides strict bandwidth guaran‑
2 4 6 8 10 12 14 16 18 20
Stream Mean Rate π (Streams/Second) tees to ST lows without any low or congestion control
mechanism at the source. Using an automated network
Fig. 21 – Decentralized Bidirectional Topology: Maximum ST packet de‑ con iguration is an imperative tool set to provide a uni‑
lay for TAS.
ied communication platform based on commercial of the
17 shelf (COTS) full‑duplex Ethernet with high bandwidth
and low complexity compared to Controller Area Net‑
16.5 works (CANs), Local Interconnect Networks (LINs), and
Mean Signaling Delay (υs) 15.5 Reconfiguration - ST τ = 2 cations (e.g., industrial control, automotive, and avionics).
16
specialized ield‑buses in industrial control system appli‑
Network designs based on the IEEE 802.1Qcc framework
15
and the 802.1Qbv traf ic shaper can form a contract with
Reconfiguration - ST τ = 3
the source to forward mission critical traf ic and to auto‑
14.5
Reconfiguration - ST τ = 4
Reconfiguration - ST τ = 5
No Reconfiguration - ST τ = 2
14
No Reconfiguration - ST τ = 3
for the full lifetime of the stream. Additionally, depending
No Reconfiguration - ST τ = 4 mate the network con iguration process using 802.1Qcc
No Reconfiguration - ST τ = 5
13.5 on the forwarding plane port traf ic shaper (e.g., TAS), the
2 4 6 8 10 12 14 16 18 20
Stream Mean Rate π (Streams/Second) required schedules can be passed to the switch servers
using general user/network information protocols (e.g.,
Fig. 22 – Decentralized Bidirectional Topology: Average stream signal‑ TLV, NETCONF/Yang, and SNMP).
ing delay for TAS.
In this paper, we have investigated the impact of TAS
We found that the stream signaling overhead with the de‑ recon igurations in response to dynamic network con‑
centralized bidirectional model is similar to the decen‑ ditions, i.e., the addition and removal of transient ST
tralized unidirectional model (cf. Fig. 20), albeit slightly streams ( lows) with different lifetimes. We have demon‑
lower due to the shorter signaling hop counts in the bidi‑ strated the effectiveness of TAS with and without the CNC,
rectional ring network. i.e., for centralized (hybrid) vs. decentralized (fully dis‑
While omitted for space, additional evaluations have tributed) models. We have examined network QoS traf ic
found that the throughput of the bidirectional decentral‑ characteristics when admitting ST lows based on an iter‑
ized model is nearly identical to the centralized model. ative heuristic approach that computes TAS schedules for
We observed only very slightly reduced throughput with current and newly requested ST streams.
the decentralized model compared to the centralized Based on the insights from the present study we out‑
model since the decentralized model carries the control line the following future research directions. First, it
traf ic in‑band, which very slightly reduces the link uti‑ would be interesting to judiciously change the GCL time
lization for data traf ic. for switches during recon iguration whilst satisfying QoS
Similar to all the preceding models and topologies, ST requirements. The studied recon iguration techniques
streams have zero traf ic drops. The BE packet loss rates should also be examined in alternate approaches for pro‑
for the decentralized bidirectional model are nearly iden‑ vidingdeterministicQoS,e.g.,[61,81]aswellasinthecon‑
tical to the centralized bidirectional model. Similarly, the text of related QoS oriented routing approaches, e.g. [17,
overall performance is largely improved under the bidi‑ 37].
rectional topology compared to the unidirectional model Another interesting future research direction is to adapt
due to the additional port and path. the recon iguration mechanisms that have been devel‑
The decentralized model was found to operate nearly oped in this study to the interactions between TSN and
identically to the centralized model in terms of QoS met‑ ifth generation (5G) wireless communication systems
rics and overall admission rate. Thus, the segregation of that operate with Ultra‑Reliable Low‑Latency Communi‑
traf ic based on the class of service can be accomplished cation (URLLC). A few recent studies have begun to ex‑
with the proposed decentralized model without the over‑ plore the use of TSN in the 5G URLLC context, see e.g., [28,
head complexities of a CNC node. A main disadvantage of 45, 52, 71], indicating signi icant potential for improving
the decentralized model is the in‑band CDT traf ic which 5G URLLC services by exploiting TSN. The TSN recon ig‑
can delay ST streams, particularly affecting the maximum uration mechanisms developed in this study can poten‑

28 © International Telecommunication Union, 2021