Skip to content

Latest commit

 

History

History
84 lines (62 loc) · 4.11 KB

README.md

File metadata and controls

84 lines (62 loc) · 4.11 KB
Raw

Introduction

Welcome to the official source code repository for the paper In-Network Address Caching for Virtual Networks (ACM SIGCOMM, 2024).

BibTeX 
@inproceedings {10.1145/3651890.3672213,
author = {Lior Zeno, Ang Chen, and Mark Silberstein},
title = {In-Network Address Caching for Virtual Networks},
booktitle = {Proceedings of the ACM SIGCOMM 2024 Conference},
year = {2024},
address = {Sydney, NSW, Australia},
url = {https://doi.org/10.1145/3651890.3672213},
publisher = {Association for Computing Machinery},
}

This repository includes the following components:

  • SwitchV2P simulation code for NS3
  • Required patches for NS3
  • Scripts for reproducing the paper's results
  • SwitchV2P P4 prototype

Dependencies

The code has been tested with the following versions:

  • Ubuntu 22.04
  • NS 3.36.1
  • Boost 1.74
  • Z3 4.8.12

Reproducing the results

To reproduce the results from the paper, first execute the build script: ./build.sh. This script will:

  • Download the specified version of NS3
  • Install required dependencies
  • Patch NS3
  • Build the code

Running the experiments

  1. Set the NS3_HOME environment variable: source ./set-ns3-home.sh
  2. Navigate to the results directory: cd ./ns3/scratch/switchv2p/results
  3. Run the desired experiment using the auxilary script: ./run.py -w <WORKLOAD> [--gw | --topo]
  • Replace <WORKLOAD> with one of the supported options: hadoop, websearch, alibaba, microburst, video.
  • To run the gateway or topology scaling experiments, include the --gw or --topo options, respectively.
  • Edit run.py and set the value of MAX_PROC to the number of CPU cores you want to allocate for the experiments. If you encounter out-of-memory (OOM) exceptions, reduce the MAX_PROC value accordingly.
  1. Run the VM migration test (section 5.2) using the auxilary script: ./run_migration_test.sh

Understanding the results

Results for each workload are stored in separate folders. For example, the results for hadoop are stored in a folder named hadoop. Each subfolder within these folders represents a different run, and contains a config.json file with the run's configuration. The results of each run are stored in a results.json file, which includes the following keys:

  • total_tx_packets: The total number of packets sent during the simulation.
  • total_rx_packets: The total number of packets received during the simulation.
  • total_dropped_packets: The total number of packets dropped during the simulation.
  • total_gw_packets: The total number of packets processed by the gateways during the simulation.
  • total_learning_packets: The total number of learning packets generated by SwitchV2P during the simulation.
  • total_invalidation_packets: The total number of invalidation packets generated by SwitchV2P during the simulation.
  • total_misdelivered_packets: The total number of misdelivered packets during the simulation.
  • last_misdelivered_packet: The timestamp when the last misdelivered packet was received by the previous destination.
  • switch_to_processed_bytes: A mapping between switch IDs and the number of bytes they processed during the simulation. In the FT8-10K topology, core switches are 0-15, spines are 16-47, and ToRs are 48-79.
  • switch_to_hits, switch_to_first_hits: Mappings between switch IDs and the number of cache hits for any packet and the number of cache hits for first packets in a flow during the simulation.
  • avg_fpl: The average first packet latency in the simulation.
  • avg_fct: The average flow completion time in the simulation.
  • avg_packet_latency: The average latency of packets during the simulation.
  • avg_packet_hops: The average number of hops each packet took during the simulation.

To extend the reported metrics, you can modify the trace-sim.cc file located under scratch/switchv2p. For example, to report the number of packets each switch processed during the simulation, see line 137.

License

This code is licensed under the MIT License. See the LICENSE file for more details.