Opened 6 years ago
Closed 6 years ago
#10460 closed enhancement (implemented)
Graph advertised bandwidth of 100th fastest relay and 100th fastest exit
Reported by: | karsten | Owned by: | karsten |
---|---|---|---|
Priority: | Medium | Milestone: | |
Component: | Metrics/Website | Version: | |
Severity: | Keywords: | ||
Cc: | arma | Actual Points: | |
Parent ID: | Points: | ||
Reviewer: | Sponsor: |
Description
Roger tells me the reason why he's looking at the fast-exits graphs on the metrics website is to get a handle on whether we're maintaining a good population of fast exits. Each time he looks at them he wishes he had a different graph though. The graph he wants to see has to do with growth of the network over time in terms of number of good relays. He says maybe the graph he wants is something like "the speed of the 100th fastest relay" and "the speed of the 100th fastest exit". And he says by 'speed' he probably means self-advertised capacity.
Let's discuss this new graph on this ticket. If we come up with something we like, we can replace the fast-exits graphs on the metrics website with the new graph.
Child Tickets
Attachments (2)
Change History (4)
Changed 6 years ago by
Attachment: | advbw-nth-fastest-relay-10-20-50-100-200-500-1000-2000.png added |
---|
Changed 6 years ago by
Attachment: | advbw-percentiles-100-99-98-97.png added |
---|
Advertised bandwidth percentiles
comment:1 Changed 6 years ago by
Status: | new → needs_information |
---|
I made two graphs:
The first graph shows the advertised bandwidth of the nth-fastest relay in the network. You asked for the 100th fastest relay or exit, but I added a few more n's from which we can pick. We can kick out all n's but one for the final graph if that makes it easier to understand.
What I don't like about this first graph is that n=100 is an absolute number that doesn't grow or shrink with the number of considered relays. There are roughly 5000 relays in the network today, so the 100th fastest is in the top 2%. But there are only about 1000 exits in the network, so the 100th fastest exit is only in the top 10%. This gets worse if you look at the network over time, which only had 1000 relays in 2007, but keep the number 100 fixed. It would be better to pick a fixed percentage value.
That's why I made a second graph with advertised bandwidth percentiles. It shows the 100-th percentile which is simply the maximum advertised bandwidth, and the 99th, 98th, and 97th advertised bandwidth percentiles. For example, the 98th percentile value is the advertised bandwidth that 98% of relays don't exceed and that 2% of relays don't fall below. With 5000 relays, it's roughly the 100th fastest relay in the network. With 1000 exits, the 98th percentile is the 20th fastest exit in the network.
However, I'm not sure which graph makes more sense to people. Or maybe there's a third variant that is even easier to understand? Thoughts?
comment:2 Changed 6 years ago by
Resolution: | → implemented |
---|---|
Status: | needs_information → closed |
There are now two new graphs on the metrics website:
- https://metrics.torproject.org/network.html#advbwdist-perc
- https://metrics.torproject.org/network.html#advbwdist-relay
I also removed the fast-exit graphs.
That concludes this ticket. Closing.
Advertised bandwidth of nth-fastest relay