wiki:org/meetings/2017Montreal/Notes/MetricsRoadmap

Facilitator: Karsten Loesing

  • Metrics team met in Berlin to plan out roadmap for next 12 months
    • karsten, iwakeh, irl
    • users of data
      • operators want to understand how the network is operation
      • journalists trying to understand social impact
      • developers want to use data
    • have done refactoring to make metrics improvements easier
    • metrics team runs services
      • CollecTor
      • ExoneraTor
      • Atlas
    • categorized metrics work
      • services
      • development
      • news (e.g. journalists)
      • code (e.g. refactoring)
      • about (what does the team do, putting information on the website)
      • operation
      • internal team processes
      • data sources, improving their documentation
      • research (e.g. interfacing with researchers)
      • visualization
  • ideas for things to work on that may not have been come up in team meeting
    • metrics team is currently overscheduled
    • include tasks to remove
    • metrics could be providing more data
      • onion service data
        • number of connections
        • number of lookups
        • number of onion services that ever get visited
        • longevity of services
      • client information
        • what ASes
        • overall client churn rates
      • exit information
        • what kinds of websites are being visited (roughly, say, correlation with Alexa lists)
        • what kinds of applications seem to be popular (Web, file sharing, etc.)
      • relay operations data
        • CPU utilization
        • available file descriptors
        • connectivity
      • network performance data
        • bandwidth
        • latency
        • packet-loss rates
      • BGP announcements
        • look for anomalous route announcements (BGP hijacks)
        • data sources could include BGPMon, Route Views, RIPE RIS
      • annotate metrics graphs with news events
      • measure blockage of Tor
        • which major sites are unreachable, CAPTCHAd, or otherwise modified
        • which exits are blocked and when
        • which CDNs are engaged in blocking
      • measuring abuse complaints to exit relay operators
      • add external data - a portal for Tor-related data
        • OnionPerf data
        • BGP data
    • documentation of statistics
    • relay operators could use a simple tool to determine if relay is operating
      • Atlas already provides data on relay status on an hourly basis
      • instantaneous and network-wide measurement would still be helpful
      • could be done with automated construction of test two-hop circuits
    • adding visualizations
      • relay population churn
        • could help for botnet detection
        • per-ASN breakdown
      • GitHub commits to show development activity
        • capture developer activity outside of tor master branch
      • Tor health graphics based on OONI measurements
        • which pluggable transports are being blocked and where?
      • Tor performance based on region (different OnionPerf instances see different performance)
    • operations
      • scanning the network for misbehavior
        • information on undeclared families (aka Sybil hunting)
        • bad exit scanning
        • bad HSDir detection
        • can this be published without compromising the operations?
        • could give confidence that Tor is protecting its network
        • could also give a bad impression that Tor is unsafe
        • is anybody storing historical information about relays that are being kicked out?
    • news
      • set up Twitter bot
        • report events in the world that affect Tor
        • report events interesting events within Tor
          • developer activity, Tor traffic activity, change in relay population
Last modified 3 months ago Last modified on Oct 30, 2017, 11:59:19 PM