Time-Aware Shaping - Showcase

[adamgeorge309]

Discuss the "Time-Aware Shaping" showcase here. Ideas and recommendations are welcome.

[ar8384]

Hello everyone,

I have trouble to understand the logic behind the offset parameter of PeriodicGate.
As far as I understand, the real value of the offset is offset - duration while offset > 0 and offset > duration. In which case, a configuration where offset = 8ms and duration = [2ms, 2ms] would be equivalent to offset = 0ms and duration = [2ms, 2ms]. Am I right?

Thank you in advance

(I also saw what seems to be a misprint to me in this line: *.client.app[*].source.packetLength = 1000B - 54B # 42B = 8B (UDP) + 20B (IP) + 14B (ETH MAC) + 4B (ETH FCS) + 8B (ETH PHY): shouldn't it be ... # 54B = ... ?)

[adamgeorge309]

Yes, you are right on both; offset = 8ms and durations = [2ms, 2ms] is equivalent to offset = 0ms, because the offset is an integer multiple of the duration. So you just shift the closing and opening times with two complete gate cycles.

So basically the durations parameter specifies when the gate is open and closed within a duration of a gate cycle. This pattern then repeats in each gate cycle. The offset parameter can shift this within the gate cycle. So it is possible to describe the same pattern with just using the durations parameter.

The offset shifts the pattern by selecting a point in the pattern, and the pattern starts at that point in the beginning of the gate cycle.

Also, thanks for the typo, it is 54B

[Kueshen]

Hi,
I am following the example posted here https://inet.omnetpp.org/docs/showcases/tsn/trafficshaping/timeawareshaper/doc/index.html. However, I cannot find the NED file used to generate the network. Any chance you can upload this file?

Regards

[adamgeorge309]

It is using TsnLinearNetwork (in the ini file: network = inet.networks.tsn.TsnLinearNetwork). It is located in inet/src/networks/tsn/TsnLinearNetwork.ned.

[Kueshen]

Thanks for the info! Do you have any example of 802.1AS and 802.1Qbv working together? I have issues when enabling both. See example attached. Any clue? Thanks!

Ned file:
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with this program. If not, see http://www.gnu.org/licenses/.
//

package simple_qbv;

@license(LGPL);

//package inet.examples.ieee8021as;
import inet.networks.base.TsnNetworkBase;
import inet.linklayer.ieee8021as.GptpMaster;
//import inet.linklayer.ieee8021as.GptpBridge;
//import inet.linklayer.ieee8021as.GptpMaster;
//import inet.linklayer.ieee8021as.GptpSlave;
//import inet.networklayer.configurator.contract.IL3NetworkConfigurator;
//import inet.networklayer.configurator.contract.INetworkConfigurator;
//import inet.networklayer.configurator.ipv4.Ipv4NetworkConfigurator;
//import inet.node.ethernet.Eth100M;
import inet.node.ethernet.EthernetLink;
import inet.node.tsn.TsnClock;
//import inet.node.inet.StandardHost;
import inet.node.tsn.TsnDevice;
import inet.node.tsn.TsnSwitch;
import inet.common.scenario.ScenarioManager;
//import inet.visualizer.common.GateScheduleVisualizer;

network SimpleReferenceQbv extends TsnNetworkBase
{
@display("bgb=1127.74,565");
types:
channel Ethernet100 extends ned.DatarateChannel
{
datarate = 100Mbps;
delay = 25ns; // Length of connection is 5 meters
}
submodules:
client: TsnDevice {
@display("p=900,200");
}
server: TsnDevice {
@display("p=400,200");
}
switch4: TsnSwitch {
@display("p=800,200");
}
switch3: TsnSwitch {
@display("p=700,200");
}
switch2: TsnSwitch {
@display("p=600,200");
}
switch1: TsnSwitch {
@display("p=500,200");
}
grand_leader: TsnClock {
@display("p=500,75");
}
connections:
client.ethg++ <--> Ethernet100 <--> switch4.ethg++;
switch4.ethg++ <--> Ethernet100 <--> switch3.ethg++;
switch3.ethg++ <--> Ethernet100 <--> switch2.ethg++;
switch2.ethg++ <--> Ethernet100 <--> switch1.ethg++;
switch1.ethg++ <--> Ethernet100 <--> server.ethg++;
grand_leader.ethg++ <--> Ethernet100 <--> switch1.ethg++;
}

omnetpp.ini

[General]
network = simple_qbv.SimpleReferenceQbv
sim-time-limit = 1s
description = "Traffic shaping using time-aware shapers + 802.1AS"

*.visualizer.typename = "IntegratedMultiCanvasVisualizer"
*.visualizer.infoVisualizer.displayInfos = true
#**.displayGateSchedules = true

Set all reference clocks to master clock so the time difference can be visualized

**.referenceClock = "grand_leader.clock"

enable time synchronization in all network nodes

..hasTimeSynchronization = true

data link visualizer displays gPTP time synchronization packets

*.visualizer.dataLinkVisualizer[0].displayLinks = true
*.visualizer.dataLinkVisualizer[0].activityLevel = "protocol"
*.visualizer.dataLinkVisualizer[0].packetFilter = "GptpSync"
*.visualizer.dataLinkVisualizer[0].lineColor = "blue2"

*.visualizer.numInfoVisualizers = 3
.visualizer.infoVisualizer[0].modules = ".grand_leader.clock"
.grand_leader.clock.displayStringTextFormat = "time: %T"
.visualizer.infoVisualizer[1].modules = ".switch.clock"
.visualizer.infoVisualizer[1].placementHint = "topLeft"
.visualizer.infoVisualizer[2].modules = ".tsnDevice.clock"
*.visualizer.infoVisualizer[2].placementHint = "bottom"
.switch.clock.displayStringTextFormat = "diff: %d"
*.server.clock.displayStringTextFormat = "diff: %d"
*.client.clock.displayStringTextFormat = "diff: %d"

.gateFilter = ".eth[1].**"
**.gateScheduleVisualizer.height = 16
**.gateScheduleVisualizer.placementHint = "top"

all oscillators have a random constant drift

**.oscillator.typename = "ConstantDriftOscillator"
**.oscillator.driftRate = uniform(-100ppm, 100ppm)

all Ethernet interfaces have 100 Mbps speed

..eth[*].bitrate = 100Mbps

client applications

*.client.numApps = 2
.client.app[].typename = "UdpSourceApp"
*.client.app[0].display-name = "best effort"
*.client.app[1].display-name = "video"
.client.app[].io.destAddress = "server"
*.client.app[0].io.destPort = 1000
*.client.app[1].io.destPort = 1001
.client.app[].source.packetLength = 1000B - 54B # 42B = 8B (UDP) + 20B (IP) + 14B (ETH MAC) + 4B (ETH FCS) + 8B (ETH PHY)
*.client.app[0].source.productionInterval = exponential(200us) # ~40Mbps
*.client.app[1].source.productionInterval = exponential(400us) # ~20Mbps

server applications

*.server.numApps = 2
.server.app[].typename = "UdpSinkApp"
*.server.app[0].io.localPort = 1000
*.server.app[1].io.localPort = 1001

enable outgoing streams

*.client.hasOutgoingStreams = true

client stream identification

*.client.bridging.streamIdentifier.identifier.mapping = [{stream: "best effort", packetFilter: expr(udp.destPort == 1000)},
{stream: "video", packetFilter: expr(udp.destPort == 1001)}]

client stream encoding

*.client.bridging.streamCoder.encoder.mapping = [{stream: "best effort", pcp: 0},
{stream: "video", pcp: 4}]

enable egress traffic shaping

*.switch4.hasEgressTrafficShaping = true
*.switch3.hasEgressTrafficShaping = true
*.switch2.hasEgressTrafficShaping = true
*.switch1.hasEgressTrafficShaping = true

time-aware traffic shaping

.switch4.eth[].macLayer.queue.numTrafficClasses = 2
.switch4.eth[].macLayer.queue.[0].display-name = "best effort"
.switch4.eth[].macLayer.queue.[1].display-name = "video"
.switch4.eth[].macLayer.queue.transmissionGate[0].offset = 0ms
.switch4.eth[].macLayer.queue.transmissionGate[0].durations = [4ms, 6ms] # period is 10 # length of periods
.switch4.eth[].macLayer.queue.transmissionGate[1].offset = 6ms
.switch4.eth[].macLayer.queue.transmissionGate[1].durations = [2ms, 8ms]

.switch3.eth[].macLayer.queue.numTrafficClasses = 2
.switch3.eth[].macLayer.queue.[0].display-name = "best effort"
.switch3.eth[].macLayer.queue.[1].display-name = "video"
.switch3.eth[].macLayer.queue.transmissionGate[0].offset = 0.1ms
.switch3.eth[].macLayer.queue.transmissionGate[0].durations = [4ms, 6ms] # period is 10 # length of periods
.switch3.eth[].macLayer.queue.transmissionGate[1].offset = 6.1ms
.switch3.eth[].macLayer.queue.transmissionGate[1].durations = [2ms, 8ms]

.switch2.eth[].macLayer.queue.numTrafficClasses = 2
.switch2.eth[].macLayer.queue.[0].display-name = "best effort"
.switch2.eth[].macLayer.queue.[1].display-name = "video"
.switch2.eth[].macLayer.queue.transmissionGate[0].offset = 0.2ms
.switch2.eth[].macLayer.queue.transmissionGate[0].durations = [4ms, 6ms] # period is 10 # length of periods
.switch2.eth[].macLayer.queue.transmissionGate[1].offset = 6.2ms
.switch2.eth[].macLayer.queue.transmissionGate[1].durations = [2ms, 8ms]

.switch1.eth[].macLayer.queue.numTrafficClasses = 2
.switch1.eth[].macLayer.queue.[0].display-name = "best effort"
.switch1.eth[].macLayer.queue.[1].display-name = "video"
.switch1.eth[].macLayer.queue.transmissionGate[0].offset = 0.3ms
.switch1.eth[].macLayer.queue.transmissionGate[0].durations = [4ms, 6ms] # period is 10 # length of periods
.switch1.eth[].macLayer.queue.transmissionGate[1].offset = 6.3ms
.switch1.eth[].macLayer.queue.transmissionGate[1].durations = [2ms, 8ms]

time sync set up

*.grand_leader.gptp.masterPorts = ["eth0"]

*.server.gptp.slavePort= "eth0"
*.client.gptp.slavePort= "eth0"

*.switch1.gptp.slavePort= "eth0"
*.switch1.gptp.masterPorts = ["eth1"]
*.switch2.gptp.slavePort= "eth0"
*.switch2.gptp.masterPorts = ["eth1"]
*.switch3.gptp.slavePort= "eth0"
*.switch3.gptp.masterPorts = ["eth1"]
*.switch4.gptp.slavePort = "eth0"
*.switch4.gptp.masterPorts = ["eth1"]

[levy]

Did you see this example?

https://inet.omnetpp.org/docs/showcases/tsn/combiningfeatures/gptpandtas/doc/index.html

[wehyy]

Hello everyone,
May I ask a question about the OMNeT++ sequence charts? I would like to know how to modify the message color in the OMNeT++ ELOG file, similar to the example in this showcase.
this is in the showcase:

and this is in my trying:

[levy]

You can do this on the Sequence Chart in the IDE only. There's a color gradient icon on the toolbar that activates the color coding dialog. You can assign colors to message filter expressions in that dialog. I hope this helps.

[levy]

Could you please make a new issue for this? As I see, it isn't directly related to this showcase.

[ar8384]

Yes, you're right. I'll do it.

[tuantitu22]

Hello,
When I try to add numTrafficClasses to 8, and allocate 8 apps to 8 pcps. Then stream of pcp 1 is transmitted when transmissionGate 0 is open, and stream of pcp 0 is transmitted when transmissionGate 1 is open. The others are fine. Is it correct? Why there is a reverse for gate 0 and 1?
Thanks!

[levy]

That is correct. See PcpTrafficClassClassifier.ned file, there's an excerpt from the IEEE Standard for Local and Metropolitan Area Networks, Bridges and Bridged Networks, Table 8-5—Recommended priority to traffic class mappings.

[tuantitu22]

Get it. Thank you very much!

[Abh4git]

Hi,

Can any of you help me with the buffer size of a TSN Switch by default? My links are Eth100M. I couldn't find this detail anywhere
Thank you!

Thanks
Regards
Abhilash

[levy]

The PacketQueue module doesn't have default limits but you can configure them using packetCapacity and dataCapacity parameters.

[Abh4git]

Thank you!

[Abh4git]

In Time Aware shaping showcase we could work with the TSNScheduleConfigurator and Scheduler. Is there a similar schedule for EDF (Earliest Deadline First) ? Is EagerScheduler EDF based? or is it just greedily scheduling. If not there can you please suggest me a path to have an EDF Scheduler? Thanks a lot for the help!

[levy]

The EagerGateScheduleConfigurator is just a greedy scheduler, it can easily fail to find a solution. I don't really know what an EDF scheduler is, but you can always implement your own GateScheduleConfigurator module.

[Abh4git]

Thank you very much! So that means the Earliest Deadline First scheduler needs to be created. Thank you for providing clarity.

[madhusrivatsav]

Can you give me some reference on how to implement our own Gateschedule configurator ?

[Abh4git]

Yes Better to have your own Base as well

1. Create your own implementation of TestGateScheduleConfiguratorBase : public NetworkConfiguratorBase
2. Now we can create our scheduler MyTestGateScheduleConfigurator : public TestGateScheduleConfiguratorBase
3. We can use these in our configuration files.
   Note: The code for existing schedulers are in inet/src

[madhusrivatsav]

Thank you for your reply. I ll check it out

[madhusrivatsav]

Is there any example that uses the gate control list to control the gates in a TSN network ?

[Abh4git]

Yes, Its TSN Gatescheduling, right there in the showcase folder with inet. https://inet.omnetpp.org/docs/showcases/tsn/gatescheduling/tsnsched/doc/index.html

[madhusrivatsav]

Thank you for your reply.
I want to create my own gate schedule to control the gates, and not use TSNschedGateScheduleConfigurator. Is there any example that shows that ?

[Frt0214]

Hi mate! I had a similar problem. Do you have any ideas about the example of configuring gating entries yourself?

[Abh4git]

Hi, The code is the best example. There are EagerGateschedulerconfigurator as well. You can see how they have implemented and go ahead with your own. RegardsAbhilash  On Tuesday, January 30, 2024 at 04:03:14 PM GMT+5:30, Madhuvanthi Srivatsav R ***@***.***> wrote: Thank you for your reply. I want to create my own gate schedule to control the gates, and not use TSNschedGateScheduleConfigurator. Is there any example that shows that ? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[madhusrivatsav]

Hi .. I am not able to run TSNSched simulation on windows. I have installed TSNSched and Z3 solver. Can you please help ?

[Abh4git]

Hi Please explain with exact environment and what problems you are facing including the steps and error details. Then only someone can help.  On Tuesday, January 30, 2024 at 09:18:41 PM GMT+5:30, Madhuvanthi Srivatsav R ***@***.***> wrote: Hi .. I am not able to run TSNSched simulation on windows. I have installed TSNSched and Z3 solver. Can you please help ? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[madhusrivatsav]

Hi
Thanks for your reply. The issue got resolved.

[madhusrivatsav]

Is it possible to set PCP values using an expression?
I am trying to set it like this.
*.client.bridging.streamCoder.encoder.mapping = [{stream: "R1", pcp: expr(int(simTime()*1000))}]

[Frt0214]

Hello everyone!

I have trouble to config my own gcl. Due to the propagation delay of links and the transmission delay of switching nodes, packets do not arrive gated immediately. After I set the hasGcl parameter to true, how do I configure some parameters so that the gate entry starts running GCL after the propagation delay of the link and the transfer delay of the switch node? This is like starting the Gcl run when Operbasetime>=Currenttime is specified in IEEE802.1Qbv. Here is my gating configuration, I want to start running Gcl entries after 8936ns.

*.switch1.eth[3].macLayer.queue.numTrafficClasses = 3
*.switch1.eth[3].macLayer.queue.hasGcl=true
*.switch1.eth[3].macLayer.queue.gateControlList.numGates=3
*.switch1.eth[3].macLayer.queue.gateControlList.offset = 8936us
*.switch1.eth[3].macLayer.queue.gateControlList.gateStates=["001","110","000"]
*.switch1.eth[3].macLayer.queue.gateControlList.durations = [8us,984us,8us]
*.switch1.eth[3].macLayer.queue.gateControlList.mapping="012"
*.switch1.eth[3].macLayer.queue.[0].display-name = "interface"
*.switch1.eth[3].macLayer.queue.[1].display-name = "be"
*.switch1.eth[3].macLayer.queue.[2].display-name = "tt"

Best Regards
Jack

[Prabharoop]

Hi,
I am working on TSN. Just a basic example like in the showcase of 2 TSN end devices and 1 switch in between. I would like to see the number of packets received at the switch and leaving the switch on a real-time basis. I noticed that qt gui shows some of this information. Can this information be collected outside and displayed as real-time data while simulations are running? Can you please help us with any pointers or examples in this direction?
@levy @avarga

[adamgeorge309]

The StatisticVisualizer module can display statistics in real-time in Qtenv. More info here: https://inet.omnetpp.org/docs/showcases/visualizer/canvas/statistic/doc/index.html

[Prabharoop]

Hi,

Thank you for your answer. But as mentioned, I need to collect these signals/ statistics in real-time, possibly from another module (For example, a cSimpleModule?) so that we can use it for further processing in real-time, rather than monitoring from Qtenv.

[anoikob]

Hi, I am trying to visualize some gates, as explained (kinda poorly, because of the lack of explicit syntax )here:

I managed to get one gate to work, by using .gateFilter = "*.tsnSwitch1.eth[9]." :

However, I want to do the same to the rest of the switches, as i want to see their TAS gates as well. What syntax am i supposed to use? I basically want more filters, but using the same command with different switch and eth ports gives me either a parse error or a duplicate line error.

[levy]

You can use wildcards and OR, AND, NOT boolean operators. For example: *.tsnSwitch*.eth[*] or *.tsnSwitch1.eth[0] or *.tsnSwitch2.eth[0].

[anoikob]

Oh thanks a lot, I will try that!

Edit: It worked. With the command .gateFilter = "*.tsnSwitch1.eth[9]. OR *.tsnSwitch3.eth[5..6].**" I can see the specified gates .

Now I just need to place them differently so that they are more readable. Thanks a lot levy!

[anoikob]

Hello and happy new year! I would like some more clarification about the PCP classifier, as I don't understand something. According to the ned file, "The
PCP is determined by a PcpReq or a PcpInd or both" . However, in the showcase, the pcp is explicitly inserted :

I am trying to create a simulation with 2 time critical traffic streams and 1 best effort stream. Which PCP values am I supposed to insert to make that happen? How do I find said values? Is there like an "algorithm" I can follow to find the proper pcp values?

[Abh4git]

Happy new year!
PCP can hold values 0 to 7. 7 being highest priority. So time critical streams can be say 6 or 7. Add another mapping after comma , {stream: "my stream", pcp: 6} Example below

*.ES1.hasOutgoingStreams = true

client stream identification

*.ES1.bridging.streamIdentifier.identifier.mapping = [{stream: "best effort", packetFilter: expr(has(udp) && udp.destPort == 1004)},
{stream: "video", packetFilter: expr(has(udp) && udp.destPort == 1002)}]

client stream encoding

*.ES1.bridging.streamCoder.encoder.mapping = [{stream: "best effort", pcp: 0},
{stream: "video", pcp: 4}]

[anoikob]

I know what values it can hold, but as far as I've understood, the pcp number together with the number of traffic classes are used to determine the index of the gate the traffic will be directed to. My question is how do I compute the pcp manually and correctly?

[Abh4git]

Oh ok, as far as I understand, the PCP class is defined based on the application.

[Kostas-Tsak]

Hello, i am fairly new at OMNeT and I wanted to ask how can I measure the latency and jitter on the TAS showcase example. Running the simulation and looking at the browse data I don't see any parameters that can give me those measurements (like End-to-end delay, packet delay, etc).

[anoikob]

Hey there! The latency and jitter metrics are already present in the simulation. All you have to do is run it once, then check the "results" folder under the simulation name. You will see 3 files: a .vec file, a .sca file and a .vci file. Opening any one of them will prompt you to create a .anf file, with which the results can be displayed.

Like so:

This is the .anf file that is created:

Double clicking the .anf file will take you to this window (if it doesn't just click the "browse data" tab shown at the bottom of the picture):

The jitter and delay metrics are both vectors, so search for them on the vectors tab.
Hope this helped!

[Kostas-Tsak]

Thanks for the very fast reply!
I have a few questions that might not fit exactly for this subject but what I found is those two parameters.
PacketDelayDifferenceToMean:vector
&
PacketDelayVariation:vector

From my understanding, PacketDelayDifferenceToMean is the delay by comparing each packet to the overall average and PacketDelayVariation is delay between consecutive packets (aka jitter). Is my understanding correct on those two parameters?
Lastly, those two parameters are taken from the server side and if yes how can I understand if it's from the server or from the switch?

[anoikob]

In the same window as the vectors, there is a "modules" tab. There you can see exactly which module is responsible for which metric, and you can further isolate them using the module filter on the top of the page. The filter's syntax is not straightforward though, but if for example you type "switch", it will show you the relevant modules and submodules. If it shows you nothing, then you need to try something else. For jitter and packet delay variation though, the metrics for those lie in the "*.sink" modules.

[Abh4git]

Just to add some details
packetDelayVariation is jitter
https://inet.omnetpp.org/docs/showcases/measurement/jitter/doc/index.html
End to End delay is meanBitLifeTimePerPacket
https://inet.omnetpp.org/docs/showcases/measurement/endtoenddelay/doc/index.html

[Kostas-Tsak]

Hello, could somebody assist me in finding the ned file. I am trying to increase the hops and run some simulations to see its effect on the network but I can't find the ned file. I tried to click on the inet4.5 folder ctrl+h and search for TsnLinearNetwork.ned but nothing was found.
Any help would be very appreciated.

[Kostas-Tsak]

I have checked the configurations of both switches in the .ini file, and they are identical. Both switches have the same gate schedules, the same transmission selection algorithm and the same PCP mappings for traffic classification.
I have also tested adjustments to the gate scheduling offsets and durations, verified that the packets arrive correctly at Switch 2 and ensured that time synchronization is active on both switches. Despite all this, the issue persists.
Would you have any suggestions for additional things I could check, especially related to internal queue processing at Switch 2? Any insight would be greatly appreciated!

[anoikob]

Ok, drop the configuration here and I'll see if I can be of help. I'm still an amateur at omnet, but others may be able to help as well.

[Kostas-Tsak]

Thanks for offering to help! Below is the configuration I am using for both the .ini and .ned files.

`//
// SPDX-License-Identifier: LGPL-3.0-or-later
//

package inet.networks.tsn;

import inet.networks.base.TsnNetworkBase;
import inet.node.contract.IEthernetNetworkNode;
import inet.node.ethernet.EthernetLink;

network TsnLinearNetwork_TwoHop extends TsnNetworkBase
{
parameters:
.eth[].bitrate = default(100Mbps);
submodules:
client: <default("TsnDevice")> like IEthernetNetworkNode {
@display("p=350,200");
}
switch1: <default("TsnSwitch")> like IEthernetNetworkNode { // First switch
@display("p=550,200");
}
switch2: <default("TsnSwitch")> like IEthernetNetworkNode { // Second switch
@display("p=750,200");
}
server: <default("TsnDevice")> like IEthernetNetworkNode {
@display("p=950,200");
}
connections:
client.ethg++ <--> EthernetLink <--> switch1.ethg++;
switch1.ethg++ <--> EthernetLink <--> switch2.ethg++; // Extra hop added
switch2.ethg++ <--> EthernetLink <--> server.ethg++;
}
[General]
network = inet.networks.tsn.TsnLinearNetwork_TwoHop
sim-time-limit = 1s
description = "Traffic shaping using time-aware shapers with two hops"

.displayGateSchedules = true
.gateFilter = ".eth[1]."
**.gateScheduleVisualizer.height = 16
**.gateScheduleVisualizer.placementHint = "top"

client applications

*.client.numApps = 2
.client.app[].typename = "UdpSourceApp"
*.client.app[0].display-name = "best effort"
*.client.app[1].display-name = "video"
.client.app[].io.destAddress = "server"
*.client.app[0].io.destPort = 1000
*.client.app[1].io.destPort = 1001
.client.app[].source.packetLength = 1000B - 54B # 54B = Ethernet overhead
*.client.app[0].source.productionInterval = exponential(200us) # Best effort traffic
*.client.app[1].source.productionInterval = exponential(400us) # Video traffic

server applications

*.server.numApps = 2
.server.app[].typename = "UdpSinkApp"
*.server.app[0].io.localPort = 1000
*.server.app[1].io.localPort = 1001

enable outgoing streams

*.client.hasOutgoingStreams = true

client stream identification

*.client.bridging.streamIdentifier.identifier.mapping = [{stream: "best effort", packetFilter: expr(udp.destPort == 1000)},
{stream: "video", packetFilter: expr(udp.destPort == 1001)}]

client stream encoding

*.client.bridging.streamCoder.encoder.mapping = [{stream: "best effort", pcp: 0},
{stream: "video", pcp: 4}]
**.switch1.bridging.streamCoder.encoder.mapping = [{stream: "best effort", pcp: 0}, {stream: "video", pcp: 4}]
**.switch2.bridging.streamCoder.encoder.mapping = [{stream: "best effort", pcp: 0}, {stream: "video", pcp: 4}]

enable egress traffic shaping

*.switch1.hasEgressTrafficShaping = true
*.switch2.hasEgressTrafficShaping = true

disable forwarding IEEE 802.1Q C-Tag

*.switch1.bridging.directionReverser.reverser.excludeEncapsulationProtocols = ["ieee8021qctag"]
*.switch2.bridging.directionReverser.reverser.excludeEncapsulationProtocols = ["ieee8021qctag"]

time-aware traffic shaping (same schedules for both switches)

.switch1.eth[].macLayer.queue.numTrafficClasses = 2
.switch1.eth[].macLayer.queue.[0].display-name = "best effort"
.switch1.eth[].macLayer.queue.[1].display-name = "video"

.switch2.eth[].macLayer.queue.numTrafficClasses = 2
.switch2.eth[].macLayer.queue.[0].display-name = "best effort"
.switch2.eth[].macLayer.queue.[1].display-name = "video"

.switch1.eth[].macLayer.queue.transmissionGate[0].offset = 0ms
.switch1.eth[].macLayer.queue.transmissionGate[0].durations = [6ms, 4ms] # Best effort transmits from 0ms to 4ms, then closes

.switch1.eth[].macLayer.queue.transmissionGate[1].offset = 4ms
.switch1.eth[].macLayer.queue.transmissionGate[1].durations = [4ms, 6ms] # Video transmits from 4ms to 10ms, ensuring no overlap

.switch2.eth[].macLayer.queue.transmissionGate[0].offset = 0ms
.switch2.eth[].macLayer.queue.transmissionGate[0].durations = [6ms, 4ms]

.switch2.eth[].macLayer.queue.transmissionGate[1].offset = 4ms
.switch2.eth[].macLayer.queue.transmissionGate[1].durations = [4ms, 6ms]
`
If you notice anything off in my configuration or have any ideas about why this might be happening, I’d really appreciate the insight!

[anoikob]

Unfortunately I cannot help you, since I don't have experience in multiple sink setups. Everything I've tested always had a singular sink. I could give you a tip though. I believe it is good practice, even in single port switches / devices to explicitly state which eth port will have transmission gates. Instead of *.switch1.eth[*].macLayer.queue.* you should instead use *.switch1.eth[1].macLayer.queue.* If you are not sure which is which, you can fire up the simulation, and check it in the simulation environment by hovering over the connection with your mouse. Sorry I was not of much help.
Maybe you could try to make the network a bit different. Instead of 1 device with 2 source apps, have 2 device with one source app each, one that sends video and another that sends generic data. Same with the sinks, one of them will receive video, the other one will receive generic data. That way, you can further isolate if something goes wrong and you can pinpoint it faster.

[Kostas-Tsak]

Thank you for the suggestion! I went ahead and explicitly defined the Ethernet port for the transmission gates as you recommended, using .switch1.eth[1].macLayer.queue. instead of .switch1.eth[].macLayer.queue.*. However, after making this change, the issue still persists. The video packets at Switch 2 are still waiting in the buffer during their active transmission window and only get transmitted once the best effort window starts.

Since the gate schedules, transmission selection algorithms, and PCP mappings are identical on both switches, I suspect there might be something else affecting the queue behavior at Switch 2.

[Kostas-Tsak]

Hello, i am trying to find a way to compare TAS with a typical Ethernet Network.
Can somebody please guide me on how i can do this so i can see the difference on delay,jitter between TAS and a Typical Ethernet.

[anoikob]

I cannot guide you precisely, but I can tell you this: Check out the TAS showcase, create a simple ethernet network with the same topology and data rates for the traffic, then when you run both simulations, make sure to activate the recording of the jitter and delay (THIS version of "delay" https://inet.omnetpp.org/docs/showcases/measurement/endtoenddelay/doc/index.html) vectors.

The documentation does mention how to activate the vector recordings, but I dont remember it off the top of my head.
After your simulation is over, you can check the results with the .vec file that was created. You want to check for "jitter" and "meanBitLifetimeperpacket" or something like that. Both of those are vectors if memory serves.

[Kostas-Tsak]

Hi, thank you for the guidance!

Here’s what I’ve done so far:

I have already created a TSN setup with TAS and a typical Ethernet baseline using the same topology (3 clients sending BE + Video traffic I matched the data rates across both scenarios, ensuring realistic offered loads (around 80-90% of the link capacity).
I enabled vector recording for meanBitLifeTimePerPacket on the server to monitor end-to-end delay across all flows.
I compared the mean delays for each setup using the generated .vec and .sca files.

However, I noticed only a small difference in delay and jitter between TAS and Ethernet, even under high offered loads.

I don't know if the results are normal or i am doing something wrong.

[anoikob]

Well, the thing is, TAS is meant to have like zero jitter in the time critical flow (Im assuming video for this one). TAS needs to be configured properly not only in terms of data rates, but also gate cycles, to prioritize and serve the time critical flows, at the expense (or not, depending on the data rate) of the BG traffic.
Your topology is not that well explained though. Typically, TAS is implemented on switches, which seem to be missing from your topology, at least in the way you described it. If your topology was like 3 clients > sw1 > sw2 > 3 receivers or whatever, then you would have to enable TAS in all connected entry and exit points of the switches, as well as calculate and implement the appropriate "gate open - close" times so that only 1 gate is ever open at a time.
If your topology is just client > receiver on the other hand, then TAS won't do much as there are no intermediate devices adding possible delay.

[Kostas-Tsak]

Thank you for your detailed input.

Just to clarify, in my current setup I do have a switch between the clients and the server, and TAS is enabled on the switch, not on the end hosts. Specifically:

The topology is:

client[0..2] --> switch --> server
TAS gate control lists (open/close cycles) are configured on the switch’s egress ports with two traffic classes (BE and video with different PCP values).

Background BE traffic and video traffic are sent concurrently from all clients to saturate the link and generate contention.

I am measuring meanBitLifeTimePerPacket at the sink for video flows under load to evaluate TAS vs ETH.

My goal is to observe how TAS reduces worst-case delays for the critical video flows while BE traffic coexists, even in this relatively simple linear topology.

As i said tho TAS results don't look that great compared to a typical Ethernet so i am not sure if that's normal or not.

Thanks again for your help.

[Abh4git]

Hi, Are you having vlan tagging for the traffic abd what is the frequency of sending and schedule. These should be all correlated. ThanksRegardsAbhilash On Saturday, June 28, 2025 at 04:33:33 AM GMT+5:30, Tsak ***@***.***> wrote: Thank you for your detailed input. Just to clarify, in my current setup I do have a switch between the clients and the server, and TAS is enabled on the switch, not on the end hosts. Specifically: The topology is: client[0..2] --> switch --> server TAS gate control lists (open/close cycles) are configured on the switch’s egress ports with two traffic classes (BE and video with different PCP values). Background BE traffic and video traffic are sent concurrently from all clients to saturate the link and generate contention. I am measuring meanBitLifeTimePerPacket at the sink for video flows under load to evaluate TAS vs ETH. My goal is to observe how TAS reduces worst-case delays for the critical video flows while BE traffic coexists, even in this relatively simple linear topology. As i said tho TAS results don't look that great compared to a typical Ethernet so i am not sure if that's normal or not. Thanks again for your help. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[Abh4git]

You can go with1. Using always open gates in omnet++ inetOr2. Eagergate scheduler which is a first in first out approach These are the typical ways to compare with ethernet ThanksRegardsAbhilash On Saturday, June 28, 2025 at 01:04:11 AM GMT+5:30, Tsak ***@***.***> wrote: Hello, i am trying to find a way to compare TAS with a typical Ethernet Network. Can somebody please guide me on how i can do this so i can see the difference on delay,jitter between TAS and a Typical Ethernet. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[Abh4git]

https://inet.omnetpp.org/docs/showcases/tsn/gatescheduling/eager/doc/index.html Alqays open caseTime-aware shaping functionality can be added to a TsnSwitch by setting the hasEgressTrafficShaping parameter to true. This setting replaces the default queue with an Ieee8021qTimeAwareShaper in all Ethernet interfaces in the switch. Note Setting this parameter only adds the option of Time-aware shaping to the switch. To use it, gate schedules need to be configured. This is because, by default, the gates are always open; thus, without any configuration, the shaper works as a priority queue, where frames are prioritized by PCP. As  inhttps://inet.omnetpp.org/docs/showcases/tsn/trafficshaping/timeawareshaper/doc/index.html?highlight=always%20open%20gates On Saturday, June 28, 2025 at 08:48:33 AM GMT+5:30, Abhilash G ***@***.***> wrote: You can go with1. Using always open gates in omnet++ inetOr2. Eagergate scheduler which is a first in first out approach These are the typical ways to compare with ethernet ThanksRegardsAbhilash On Saturday, June 28, 2025 at 01:04:11 AM GMT+5:30, Tsak ***@***.***> wrote: Hello, i am trying to find a way to compare TAS with a typical Ethernet Network. Can somebody please guide me on how i can do this so i can see the difference on delay,jitter between TAS and a Typical Ethernet. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[Kostas-Tsak]

I will try to find it out when I am on my pc. Hopefully this was the issue with my results.

[Kostas-Tsak]

Hello, thank you for your previous guidance.

I tried increasing the queue capacity using:

**.macLayer.queue[].packetCapacity = 10000
**.macLayer.queue[].dataCapacity = -1b
.client.eth[].macLayer.queue.packetCapacity = 200000
.client.eth[].macLayer.queue.dataCapacity = -1b
to avoid packet drops and to ensure the queues would not limit traffic. However, I still get incorrect results for the video traffic at the sink. Even when the video traffic is supposed to get 90% of the duty cycle, the data rate remains around ~46 Mbps, while best effort shows ~9 Mbps, instead of the expected 90 Mbps for video and 10 Mbps for best effort.

Is there something else I should check, or is there another step needed to ensure TAS correctly limits traffic according to the gate schedules while allowing the video stream to utilize the expected bandwidth?

Any additional guidance would be greatly appreciated.

Thanks again for your help!

[Kostas-Tsak]

No, from the .ini file. I dont remember the exact command though.

I managed to solve some issue but i still can't find how to see what the queue capacity value is and how to change it. Could you assist me?

[anoikob]

The queue capacity is measured in packets. I believe you were correct in your increasing of the capacity, however I am unfamiliar with "dataCapacity". Are you still getting drops? If not, you are on the right track. If yes, you should investigate your network in the simulation environment (Qtenv) by double clicking inside each module, finding the queues, or even just tracking a single packet from the source to destination, so you can see exactly where and if it drops.
I have long uninstalled omnet from this pc, so I cannot guide you as accurately as I could, I'm only saying what comes to my mind from my experience with it.
One thing you can absolutely try is isolate the flows. Instead of a single sender sending two types of flows, have one sender send video, and another sender send best effort.

[Kostas-Tsak]

The queue capacity is measured in packets. I believe you were correct in your increasing of the capacity, however I am unfamiliar with "dataCapacity". Are you still getting drops? If not, you are on the right track. If yes, you should investigate your network in the simulation environment (Qtenv) by double clicking inside each module, finding the queues, or even just tracking a single packet from the source to destination, so you can see exactly where and if it drops. I have long uninstalled omnet from this pc, so I cannot guide you as accurately as I could, I'm only saying what comes to my mind from my experience with it. One thing you can absolutely try is isolate the flows. Instead of a single sender sending two types of flows, have one sender send video, and another sender send best effort.

Thank you for your previous explanations, they helped me understand how to check values inside Qtenv. I managed to find the packetCapacity and dataCapacity values, which are currently -1 and -1b respectively. From what I understood, this means there is no queue capacity constraint, so the queues should not be the direct cause of drops due to capacity. Additionally is there any other capacity i should worry about?

However, I am still noticing packet drops, but I can’t understand from where they happen (see images below for packetSent and packetReceived counts).

If you have any suggestions on what to check next, or what might typically cause this behavior in TSN measurements, it would help me greatly for aligning my methodology chapter.

Thank you again for your help.