[tor-bugs] #25854 [Metrics/Relay Search]: Improve advertised/observed bandwidth mouseover text

[Tor Bug Tracker & Wiki]

#25854: Improve advertised/observed bandwidth mouseover text
----------------------------------+------------------------------
 Reporter:  irl                   |          Owner:  metrics-team
     Type:  defect                |         Status:  new
 Priority:  Medium                |      Milestone:
Component:  Metrics/Relay Search  |        Version:
 Severity:  Normal                |     Resolution:
 Keywords:                        |  Actual Points:
Parent ID:                        |         Points:
 Reviewer:                        |        Sponsor:
----------------------------------+------------------------------

Comment (by teor):

 Relay Search is currently ambiguous:

 The volume of traffic, both incoming and outgoing, that the relay is
 willing to sustain, as configured by the operator and claimed to be
 observed from recent data transfers.

 Replying to [comment:2 irl]:
 > The problem here is that dir-spec is not particularly clear on this, so
 probably some reading of the code is needed to make sure.

 I'll open a ticket to improve dir-spec, once we know what isn't clear.

 > Someone asked in IRC whether this was the sum of the total bandwidth
 (directory fetches, relay traffic, client traffic) in and out, or whether
 this was just for traffic that is relayed, or relayed traffic + overhead
 just for relaying traffic (and then whether it's the total in+out or just
 total throughput).

 I answered a similar question for tjr in #tor-project a week or so ago. If
 anyone has the IRC backlog, they can get the full details.
 I'll try to reconstruct it here.

 This is what the spec currently says:

        Estimated bandwidth for this router, in bytes per second.  The
        "average" bandwidth is the volume per second that the OR is willing
 to
        sustain over long periods; the "burst" bandwidth is the volume that
        the OR is willing to sustain in very short intervals.  The
 "observed"
        value is an estimate of the capacity this relay can handle.  The
        relay remembers the max bandwidth sustained output over any ten
        second period in the past day, and another sustained input.  The
        "observed" value is the lesser of these two numbers.

 https://gitweb.torproject.org/torspec.git/tree/dir-spec.txt#n424

 In other words, the spec says that:
 * average bandwidth limit is the maximum bandwidth the relay is willing to
 handle over long periods
 * burst bandwidth limit is the maximum bandwidth the relay is willing to
 handle over short periods
 * observed bandwidth rate is min(in, out) where in and out are the maximum
 bandwidths sustained inbound or outbound over any recent 10 second period
 over the last day

 The spec isn't explicit about the direction of the traffic, except for
 observed bandwidth. Here is how Tor currently implements the bandwidth
 limits:
 * there is a token bucket for inbound traffic, and tor stops reading from
 the kernel once the bucket is empty
 * there is a token bucket for outbound traffic, and tor stops writing to
 the kernel once the bucket is empty

 The difference between kernel and link-level traffic causes some
 interesting issues for people who are trying to get precise bandwidths.
 See, for example, #25687. Our general advice is "don't worry about that
 level of detail".

 So there is a separate limit on both inbound and outbound traffic.
 The effective relayed rate is min(in, out), and includes some overheads,
 just like the observed bandwidth.
 We can update dir-spec to clarify this.

 The spec also doesn't specify whether the bandwidth includes overheads, so
 you should assume that it does include some overheads, and it doesn't
 include others. In particular:
 * Recent tor versions include everything over an ORPort, whether it
 originates at the relay, is relayed for a client, or is sent to a client
 in response to a BEGINDIR request. We can update the spec to clarify this.
 * I'm not sure if Tor includes DirPort traffic. We could find out what tor
 does and specify it. But I'm not sure if this level of detail is relevant.
 * As far as I recall, Tor includes all cell packing overhead, but not TLS,
 TCP, IP, or link layer overhead. But I'm not sure if this level of detail
 is relevant.

 Can you help me understand why you need to know details about overheads?
 The exact overheads that are included will change over time as we improve
 Tor's code and statistics.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/25854#comment:3>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online