Run NS-3 on Ubuntu 20.04

This is an NS-3 simulator for "CAVER: Enhancing RDMA Load Balancing by Hunting Less-Congested Paths". We add CAVER's components to ConWeave(SIGCOMM'2023)'s NS-3 simulator to simulate RDMA load balancing performance with various algorithms.

We describe how to run this repository using your local machine with ubuntu:20.04.

0. Prerequisites

We tested the simulator on Ubuntu 20.04, but latest versions of Ubuntu should also work.

Firstly, use the following command to install the required dependencies

sudo apt install build-essential python3 libgtk-3-0 bzip2 python2

For plotting, we use numpy, matplotlib, pandas and cycler for python3:

python3 -m pip install numpy matplotlib cycler pandas

1. Configure & Build

Run following commands to configure and build simulation:

wget https://www.nsnam.org/releases/ns-allinone-3.19.tar.bz2
tar -xvf ns-allinone-3.19.tar.bz2
cd ns-allinone-3.19
rm -rf ns-3.19
git clone https://github.com/CAVER-LB/CAVER-ns3.git ns-3.19
cd ns-3.19
./waf configure --build-profile=optimized
./waf

2. Run

To run the simulation, use the following command:

./autorun <topology> <load>

For example:

./autorun.sh fat_k8_100G_OS2 60

All topology files are stored in the ./config directory, along with various traffic workloads that were used to reproduce the results from our paper. We will explain these workloads in more detail later.

Each experiment will be assigned a unique ID. The results will be stored in the ./mix/output/ directory and logged in ./mix/autorun_history.txt. For example, after running:

./autorun.sh fat_k8_100G_OS2 60

the ./mix/autorun_history.txt file will include an entry like:

Thu Jan 31 14:25:16 PST 2025 TOPOLOGY:fat_k8_100G_OS2, NETLOAD=60, RUNTIME=0.03, INDEX=[caver:3,conga:4,conweave:5,hula:6,fecmp:7,]

In this case, 3, 4, 5, 6, 7 represent the experiment IDs.

To analyze the experiments, use the functions in analysis/deep_analyse.py. For example, you can call:

print(get_basic_result('3-7'))

This will return the average FCT slowdown, P99 FCT slowdown, large FCT slowdown, and small FCT slowdown for the experiments with IDs 3, 4, 5, 6, 7.

Note: Due to random factors, the results of each experiment may vary slightly.

3. Reproduce

We also provide a simpler way to reproduce the experiments in the paper. You can use functions in caver_run.py. In particular:

• You can use run_normal in caver_run.py and plot_overall_fctslowdown in analysis/deep_analyse.py to reproduce Fig 7, Fig 17 in paper.
• You can use run_caver_pathnum_ce_experiments in caver_run.py to reproduce Fig 20 in paper.
• You can use run_caver_patchoiceTimeout_experiments in caver_run.py to reproduce Table 1 in paper.
• You can use run_experiments_with_topology in caver_run.py to reproduce Fig 8 in paper.
• You can use run_with_traffic_patterns in caver_run.py to reproduce Fig 9, Fig18, and Table 4 in paper.

4. 新增功能

本功能新增了基于接收窗口的乱序报文重排支持，通过以下技术手段优化数据包处理流程：

1. 报文标记机制​​：在每个流的最后一个数据包添加布尔标志位（bool）标识报文序列的完整性
2. 固定接收窗口​​：维护64个报文的固定窗口缓冲区
3. 自适应重排策略​​：当窗口内报文乱序率达到阈值时触发重排操作