List of datasets related to networking. Useful for data-driven evaluation or machine learning approaches. Feel free to comment with updates.

networking_datasets.md

Real-world networking datasets

Please comment below if you have an update, e.g., with another networking-related dataset.

Dataset repos or tools

• Finding datasets: New Google Dataset search
• Mendeley Data: https://data.mendeley.com/datasets
• Kaggle Data: https://www.kaggle.com/datasets
• Google's M-Lab networking performance data sets: https://www.measurementlab.net/data/

Service/VNF-related data

• SNDZoo: https://sndzoo.github.io/
• Datasets from the University of Catalunya: http://knowledgedefinednetworking.org/

Traffic

Traffic traces:

• Traffic over time (in minutes, hours, or days) from ~2005 of a private ISP. Only time + traffic size, no source/destination or any other info. https://datamarket.com/data/list/?q=cat:ecd%20provider:tsdl
• Facebook traffic traces (access via FB group): https://research.fb.com/data-sharing-on-traffic-pattern-inside-facebooks-datacenter-network/
• North American Backbone network Abilene, date from 24 weeks of 5 minute (2004) averages, 12 routers (12x12 matrices): http://www.maths.adelaide.edu.au/matthew.roughan/project/traffic_matrix/ (160 MB)
• SNDlib: Library with real-world network topologies (most popular: Abilene) and sometimes traffic/service demands: http://sndlib.zib.de/
  • Under Library > Dynamic traffic there are realistic traffic traces that are dynamically changing over time. For example traffic matrix from every 5min over 6 months for the Abilene network.
  • Also for every 15min for the larger Geant network
• UMassTraceRepository
• RawDad: Real, wireless data (124 datasets). Download only for registered users (free). https://crawdad.org/about.html
• Internet traffic archive: Useful, real-world traces. http://ita.ee.lbl.gov/html/traces.html
• MAWI Working Group Traffic Archive Packet traces from WIDE backbone
• Huge traces from Google. 12000+ machines, measured over 1 month, compressed size ~41GB (also smaller traces available). https://github.com/google/cluster-data
• IP-Network traffic labeled with different apps: https://www.kaggle.com/jsrojas/ip-network-traffic-flows-labeled-with-87-apps/home
• Summarized network traffic where computers are gradually compromised by a botnet: https://www.kaggle.com/crawford/computer-network-traffic/home
• Caida packet-level traffic traces from 2016: https://www.caida.org/data/passive/passive_2016_dataset.xml
• tcpreplay simple traffic traces (smallFlows.pcap and bigFlows.pcap) for traffic generation: https://tcpreplay.appneta.com/wiki/captures.html

Traffic Generator

• Trex generates statefull and stateless traffic. Python API is provided. https://github.com/cisco-system-traffic-generator/trex-core
• Intel's traffic testbed for data plane development kit (DPDK) (using Trex): https://software.intel.com/en-us/articles/build-your-own-dpdk-traffic-generator
• MoonGen
• How to generate Netflow data from PCAP traces (more available): https://stackoverflow.com/a/34792376/2745116

Networks

Network topologies

• SNDlib: http://sndlib.zib.de/home.action
• TopologyZoo: http://topology-zoo.org/

Network delays

• Historical (hourly) ping measurements for large-scale network: https://wondernetwork.com/pings
• Datasets from the University of Catalonia: http://knowledgedefinednetworking.org/

Other

Workload

• Alibaba traces with batch workloads on thousands of machines (from 2017 and 2018): https://github.com/alibaba/clusterdata