Hi,
At torservers.net, we run some large exit relays with an "allow all except port 25" policy.
These are statistics from ARM showing exit port statistics of a fast exit running for seven hours at 30-40 MB/s:
443 HTTPS 17650 (%55) 80 HTTP 10625 (%33) 8344 739 (%2) 8333 84 51413 45 6881 Torrent 38 8337 29 993 IMAPS 23 45682 18 182 16 9001 16 4662 eMule 14 4661 12
Thus spake Moritz Bartl (moritz@torservers.net):
At torservers.net, we run some large exit relays with an "allow all except port 25" policy.
These are statistics from ARM showing exit port statistics of a fast exit running for seven hours at 30-40 MB/s:
443 HTTPS 17650 (%55) 80 HTTP 10625 (%33) 8344 739 (%2)
How long has this relay been up using that IP? Was the exit only 7 hours old?
And is this read+write, or just one direction?
Here's the read and write statistics from the ExtraInfo descriptors from a handful of the fastest default-policy and reduced-policy relays:
Default exit lumumba read 819.7M other: 66.5% 80: 22.7% 443: 5.1% 51413: 1.4% 6881: 1.2% 182: 0.8% Default exit lumumba wrote 257.7M other: 91.4% 443: 1.9% 80: 1.8% 6881: 1.6% 62686: 1.5% 51413: 1.5%
Default exit chomsky read 744.0M 80: 44.7% other: 40.1% 443: 10.0% 182: 1.1% 51413: 1.0% 4000: 0.7% Default exit chomsky wrote 164.3M other: 85.2% 443: 5.1% 80: 4.8% 51413: 2.7% 32460: 1.0% 6881: 1.0%
Default exit politkovskaja2 read 872.3M other: 70.5% 80: 19.3% 443: 4.2% 51413: 1.8% 182: 1.1% 6881: 0.9% Default exit politkovskaja2 wrote 292.4M other: 92.8% 443: 1.6% 12762: 1.4% 51413: 1.4% 80: 1.4% 6881: 1.3%
Default exit politkovskaja read 887.8M other: 70.7% 80: 19.1% 443: 4.2% 51413: 1.5% 6881: 1.0% 23877: 0.8% Default exit politkovskaja wrote 301.3M other: 93.0% 443: 1.7% 51413: 1.5% 80: 1.5% 6881: 1.3% 45682: 0.8%
Default exit rainbowwarrior read 695.0M 80: 43.9% other: 41.7% 443: 9.5% 182: 1.4% 51413: 0.9% 6881: 0.6% Default exit rainbowwarrior wrote 148.5M other: 86.9% 443: 5.3% 80: 4.7% 51413: 1.4% 6881: 1.3% 8333: 0.2%
Misc Exit bouazizi read 662.9M 80: 78.4% 443: 20.6% 8080: 0.3% other: 0.2% 22: 0.1% 8333: 0.1% Misc Exit bouazizi wrote 30.4M 443: 56.5% 80: 42.9% other: 0.2% 8080: 0.1% 22: 0.1% 8333: 0.1%
Misc Exit assk2 read 314.6M 80: 78.1% 443: 20.4% 563: 0.5% 8333: 0.4% 8080: 0.2% 3389: 0.1% Misc Exit assk2 wrote 13.3M 443: 55.1% 80: 42.9% 8333: 1.1% other: 0.9% 8080: 0.1% 3128: 0.0%
Misc Exit assk read 275.6M 80: 78.6% 443: 20.3% 8333: 0.5% 8080: 0.3% 3389: 0.1% other: 0.1% Misc Exit assk wrote 12.4M 443: 55.5% 80: 41.9% 8333: 2.0% other: 0.6% 8080: 0.1% 995: 0.0%
Misc Exit Amunet6 read 265.1M 80: 78.0% 443: 19.6% 8080: 1.0% 22: 0.9% 3389: 0.1% 8333: 0.1% Misc Exit Amunet6 wrote 10.7M 443: 54.2% 80: 44.7% 8333: 0.4% other: 0.4% 22: 0.1% 81: 0.1%
Misc Exit noiseexit01b read 250.2M 80: 78.0% 443: 20.5% 8080: 0.5% 81: 0.4% 8333: 0.2% 995: 0.1% Misc Exit noiseexit01b wrote 10.9M 443: 53.0% 80: 45.2% 8333: 1.1% other: 0.5% 8080: 0.1% 995: 0.0%
Misc Exit raskin read 670.2M 80: 77.6% 443: 21.2% 8080: 0.3% 563: 0.2% other: 0.2% 81: 0.2% Misc Exit raskin wrote 30.3M 443: 57.3% 80: 42.1% 8333: 0.3% other: 0.3% 995: 0.0% 8080: 0.0%
On Wednesday, August 15, 2012 4:44pm, "Mike Perry" mikeperry@torproject.org said: [snip]
Here's the read and write statistics from the ExtraInfo descriptors from a handful of the fastest default-policy and reduced-policy relays:
Default exit lumumba read 819.7M other: 66.5% 80: 22.7% 443: 5.1% 51413: 1.4% 6881: 1.2% 182: 0.8% Default exit lumumba wrote 257.7M other: 91.4% 443: 1.9% 80: 1.8% 6881: 1.6% 62686: 1.5% 51413: 1.5%
Default exit chomsky read 744.0M 80: 44.7% other: 40.1% 443: 10.0% 182: 1.1% 51413: 1.0% 4000: 0.7% Default exit chomsky wrote 164.3M other: 85.2% 443: 5.1% 80: 4.8% 51413: 2.7% 32460: 1.0% 6881: 1.0%
Default exit politkovskaja2 read 872.3M other: 70.5% 80: 19.3% 443: 4.2% 51413: 1.8% 182: 1.1% 6881: 0.9% Default exit politkovskaja2 wrote 292.4M other: 92.8% 443: 1.6% 12762: 1.4% 51413: 1.4% 80: 1.4% 6881: 1.3%
Default exit politkovskaja read 887.8M other: 70.7% 80: 19.1% 443: 4.2% 51413: 1.5% 6881: 1.0% 23877: 0.8% Default exit politkovskaja wrote 301.3M other: 93.0% 443: 1.7% 51413: 1.5% 80: 1.5% 6881: 1.3% 45682: 0.8%
Default exit rainbowwarrior read 695.0M 80: 43.9% other: 41.7% 443: 9.5% 182: 1.4% 51413: 0.9% 6881: 0.6% Default exit rainbowwarrior wrote 148.5M other: 86.9% 443: 5.3% 80: 4.7% 51413: 1.4% 6881: 1.3% 8333: 0.2%
Misc Exit bouazizi read 662.9M 80: 78.4% 443: 20.6% 8080: 0.3% other: 0.2% 22: 0.1% 8333: 0.1% Misc Exit bouazizi wrote 30.4M 443: 56.5% 80: 42.9% other: 0.2% 8080: 0.1% 22: 0.1% 8333: 0.1%
Misc Exit assk2 read 314.6M 80: 78.1% 443: 20.4% 563: 0.5% 8333: 0.4% 8080: 0.2% 3389: 0.1% Misc Exit assk2 wrote 13.3M 443: 55.1% 80: 42.9% 8333: 1.1% other: 0.9% 8080: 0.1% 3128: 0.0%
Misc Exit assk read 275.6M 80: 78.6% 443: 20.3% 8333: 0.5% 8080: 0.3% 3389: 0.1% other: 0.1% Misc Exit assk wrote 12.4M 443: 55.5% 80: 41.9% 8333: 2.0% other: 0.6% 8080: 0.1% 995: 0.0%
Misc Exit Amunet6 read 265.1M 80: 78.0% 443: 19.6% 8080: 1.0% 22: 0.9% 3389: 0.1% 8333: 0.1% Misc Exit Amunet6 wrote 10.7M 443: 54.2% 80: 44.7% 8333: 0.4% other: 0.4% 22: 0.1% 81: 0.1%
Misc Exit noiseexit01b read 250.2M 80: 78.0% 443: 20.5% 8080: 0.5% 81: 0.4% 8333: 0.2% 995: 0.1% Misc Exit noiseexit01b wrote 10.9M 443: 53.0% 80: 45.2% 8333: 1.1% other: 0.5% 8080: 0.1% 995: 0.0%
Misc Exit raskin read 670.2M 80: 77.6% 443: 21.2% 8080: 0.3% 563: 0.2% other: 0.2% 81: 0.2% Misc Exit raskin wrote 30.3M 443: 57.3% 80: 42.1% 8333: 0.3% other: 0.3% 995: 0.0% 8080: 0.0%
Pardon my tangent, but: The enormous discrepancy in read/write values, if accurate, makes a mockery of AccountingMax for purposes of tracking bandwidth used.
Thus spake Steve Snyder (swsnyder@snydernet.net):
On Wednesday, August 15, 2012 4:44pm, "Mike Perry" mikeperry@torproject.org said:
Here's the read and write statistics from the ExtraInfo descriptors from a handful of the fastest default-policy and reduced-policy relays:
Pardon my tangent, but: The enormous discrepancy in read/write values, if accurate, makes a mockery of AccountingMax for purposes of tracking bandwidth used.
These are per-port exit stats. I believe AccountingMax is for *all* relayed traffic.
For every byte the exit writes to a port, that is a byte it had to read on an orconn from another relay.
Similarly, for every byte an exit read from a port, that is a byte it has to write to an orconn to another relay.
Thus, for exit relays, *total* upstream and downstream will be mostly symmetric. There may be some discrepencies for packing data into 512 byte cells, though.
On Wed, Aug 15, 2012 at 10:21:18PM +0200, Moritz Bartl wrote:
At torservers.net, we run some large exit relays with an "allow all except port 25" policy.
These are statistics from ARM showing exit port statistics of a fast exit running for seven hours at 30-40 MB/s:
443 HTTPS 17650 (%55) 80 HTTP 10625 (%33) 8344 739 (%2) 8333 84 51413 45 6881 Torrent 38 8337 29 993 IMAPS 23 45682 18 182 16 9001 16 4662 eMule 14 4661 12
What's the units for the data column? Megabytes transferred over 7 hour period?
It would be interesting to see three columns: 1. number of connections 2. MB sent 3. MB received
-andy
tor-relays@lists.torproject.org