EDAF: An End-to-End Delay Analytics Framework for 5G-and-Beyond Networks

EDAF is a novel tool that decompose packets' end-to-end delays and determine each component's significance for optimizing the delay in 5G network.

To achieve that, EDAF

1. modifies the OpenAirInterface 5G protocol stack by inserting numerous time measurement points across different layers,
2. requires the application ends to report packet timestamps and sequence numbers,
3. aggregates all time measurements to analyze the end-to-end delay, and
4. generates insights for delay optimization.

Requirements and preparations

For running EDAF experiments, 3 hosts and 2 USRP SDRs are required to bring up an standalone OpenAirInterface (OAI) 5G network:

1. Core Network (CN) host
2. gNB host and SDR
3. UE host and SDR

All hosts need to be connected via a secondary out-of-band wired IP network that we call it edaf-net.

Also, their clocks must be synced by running Precision Time Protocol (PTP) clock synchronization. Therefore all hosts are supposed to be equipped with hardware timestamping capable network interface cards (NICs) which is required for PTP to sync the clocks.

Make sure Docker is installed on the hosts.

Running an experiment

1) Run 5G Core and NLMT Server

We start by running the 5G core services on CN host by following OpenAirInterface tutorials. Next, we need to run NLMT server inside the UPF or SPGWU container. In order to do that, we download the applications binary and run it using these commands:

wget https://raw.githubusercontent.com/samiemostafavi/nlmt/master/nlmt
docker cp nlmt 5gcn-spgwu:/usr/local/bin/
docker exec -it 5gcn-spgwu chmod +x /usr/local/bin/nlmt
docker exec -d 5gcn-spgwu /bin/sh -c 'while true; do nlmt server -n 172.16.32.140:50009 -i 0 -d 0 -l 0; sleep 1; done'

2) Run EDAF Server

Run EDAF on the CN host by first creating a folder on the host for storing the database.

Use the following command after creating influxdbv2 folder on the CN host:

docker run -d --rm --volume `pwd`/influxdbv2:/root/.influxdbv2 --network oai-5gcn-net --ip 172.16.32.140  --name edaf samiemostafavi/edaf:latest
docker network connect edaf-net --ip 192.168.32.140 edaf

Check influxdb UI on the browser: http://192.168.32.140:8086

3) Run 5G RAN

To run the modified gNodeB, in the config file e.g. gnb.sa.band78.fr1.106PRB.usrpb210.conf you need to add EDAF server address as:

latseq_addr = "192.168.32.140:50015";

Then run gNodeB and check EDAF logs whether it is connected or not.

For UE, when in the arguments passing to the execution command, you have to pass EDAF address as:

--latseq-addr 192.168.32.140:50011

Then run UE and check EDAF logs whether it is connected or not.

Upon successful connection between gNB and UE, run the traffic generator NLMT client on UE host to generate packets on uplink.

EDAF populates the influxDB, check the database's UI on port 8086 of CN host.

Run EDAF Offline

Instead of online networked mode, you can configure LATSEQ to produce .lseq, and NLMT to produce .json.gz files. In this case you can use offline_edaf.py script to process the data, decompose delay, and produce a parquet file. Pass the address of a folder to the script with the following structure:

FOLDER_ADDR/
-- gnb/
---- latseq.*.lseq
-- ue/
---- latseq.*.lseq
-- upf/
---- se_*.json.gz

For example:

python offline_edaf.py 240103_011728_FINAL_expB_Q1_results res.parquet

Run EDAF Standalone

If you are interested in using EDAF over an arbitrary network link, and not OpenAirInterface, follow this section. In such a scenario, you will lose the decomposition capability and you can only analyze the end-to-end delay. The hosts are required to be clock synchronized. For running standalone EDAF experiments, 2 hosts are required:

1. Server host
2. Client host

Follow the steps below to run EDAF in standalone configuration.

Create a docker network to run server services on it:

docker network create --driver=bridge --subnet=10.89.89.0/24 --ip-range=10.89.89.0/24 --gateway=10.89.89.1 edaf-net

1) Run EDAF Server

Run EDAF on the server host by first creating a folder on the host for storing the database.

Use the following command after creating influxdbv2 folder on the server host:

docker run -e STANDALONE=true -d --rm --volume `pwd`/influxdbv2:/root/.influxdbv2 --network edaf-net --ip 10.89.89.2 -p 0.0.0.0:8086:8086 --name edaf samiemostafavi/edaf:latest

Check influxdb UI on the browser: http://192.168.2.2:8086

2) Run NLMT Server

Run NLMT server

docker run -d --rm --name nlmt-server --network edaf-net --ip 10.89.89.3 -p 0.0.0.0:2112:2112 -p 0.0.0.0:2112:2112/udp samiemostafavi/nlmt /bin/sh -c 'while true; do nlmt server -n 10.89.89.2:50009 -i 0 -d 0 -l 0; sleep 1; done'

3) Run NLMT Client

On the client host, run nlmt client towards the server:

./nlmt client --tripm=oneway -i 10ms -f 5ms -g edaf1/fingolfin -l 500 -m 1 -d 5m -o d --outdir=/tmp/ 192.168.2.2

Publications

License

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this project except in compliance with the License. A copy of the license is included in the LICENSE file.