Why is my fast relay so slow to gain popularity?
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512 Hello everyone, newcomer here. I'm behind very fast connection (11.5 MB/sec down, 7.5 MB/sec up) and I thought that the Tor network could benefit from my connection, especially since it's apparently been under high load recently. Per the latest blog posts, I downloaded the beta TBB and configured it as a relay under Linux. It's been up for almost two days now, yet it's still being utilized at a very, very small fraction of it's potential. In the network map, I see that my relay has an advertised speed which is again much slower than it actually can be. To my knowledge, a web server can be put under full load right away, and distributing computing projects use the most of your computer right off the bat. Why doesn't Tor run computational and/or bandwidth tests and advertise my relay at a much more actual speed? I don't see why a fast relay has to start at the very bottom of the barrel to begin with. I can see the logic in slowly ramping up connections in testing for stability and scalability, but the progress is slow because it's continually fighting against that existing average. I can set my requested average and burst bandwidth, but AFAIK I can't affect my advertised bandwidth. Why are things this way, and is there anything I can do to speed up the utilization growth? Thanks in advance for any help you can provide, Jesse V. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) iQF8BAEBCgBmBQJSMJg3XxSAAAAAAC4AKGlzc3Vlci1mcHJAbm90YXRpb25zLm9w ZW5wZ3AuZmlmdGhob3JzZW1hbi5uZXQxMjgyMjhENjEyODQ1OTU1NzBCMjgwRkFB RDk3MzY0RkMyMEJFQzgwAAoJEK2XNk/CC+yAAekH/i2rHzwbGZMiL4g31d2JEkQR ZjOaDCe6aNkK/ORf5MhxW7B6L772BczjPkjuO0H1Oe41bSkglO78gCbLC+YQ1r68 0nc45dcztNDkQPhZVrkj7RconNrmfLXFozFLKY1Nr2mLy6RTg0a1yQ/WhpFa/qah cUmVnGtvIoDnDAhuCPIEu7SuGba0I4TkBph1atKAC/gQYyOSp9CpCqxy0J3Qaa3Y pcCkGC7oWW2I9g8UJOx0UKg1auiYMZ1QrwpJ6HdHhvhFSWnhWeggkvaOCDewC2pZ dQg7Odk/ISfkpdeCgd8hg2IQBh3kpI1kRVUvX0sJgbYCbizT45bwKCLf/XzMQHE= =gfnB -----END PGP SIGNATURE-----
On 2013-09-11 18:20 , Jesse Victors wrote:
Hello everyone, newcomer here.
I'm behind very fast connection (11.5 MB/sec down, 7.5 MB/sec up)
(Most folks would just call that 100mbit, that is if your MB is MegaByte, hence why 11.5 MiB/s would be more accurate).
thought that the Tor network could benefit from my connection,
Definitely!
especially since it's apparently been under high load recently. Per the latest blog posts, I downloaded the beta TBB and configured it as a relay under Linux. It's been up for almost two days now, yet it's still being utilized at a very, very small fraction of it's potential.
This blog post from today explains the effect and reasoning: https://blog.torproject.org/blog/lifecycle-of-a-new-relay
In the network map, I see that my relay has an advertised speed which is again much slower than it actually can be.
IMHO that label should be changed to 'measured speed' as the bwauths take care of that now.
To my knowledge, a web server can be put under full load right away, and distributing computing projects use the most of your computer right off the bat. Why doesn't Tor run computational and/or bandwidth tests and advertise my relay at a much more actual speed?
The bwauths do that, but they don't run very often.
I don't see why a fast relay has to start at the very bottom of the barrel to begin with.
Because otherwise introducing a large set of fast relays and thus hurt anonymity. (On the other side a determined adversary just waits a bit longer) Greets, Jeroen
(Most folks would just call that 100mbit, that is if your MB is MegaByte, hence why 11.5 MiB/s would be more accurate). Yes, it is megabytes/sec, according to Speedtest.net. Most sysadmins would say 91.74 mbit down, 60.23 mbit up, (which is as you say basically 100 mbit) but since the Tor measures in MiB/sec I just had Speedtest convert to that instead. Same thing either way I suppose.
This blog post from today explains the effect and reasoning: https://blog.torproject.org/blog/lifecycle-of-a-new-relay Wow, total coincidence that I asked this question on the day that the answer was posted on the blog (maybe I'm just a Jedi, who knows). That post explains a lot. I'll have to read it over several times as it's a bit technical.
Do I have to maintain an uptime of ~70 days to see fully utilization then? This relay is on a personal computer with a static IP, so it isn't on a dedicated server or anything like that. Usually my uptime is around several weeks though. I'll be on this same IP for months though.
Because otherwise introducing a large set of fast relays and thus hurt anonymity.
(On the other side a determined adversary just waits a bit longer)
Greets, Jeroen That makes sense. I didn't think of that, good one. Thanks Jeroen for pointing me in the right direction there, much appreciated.
Live long and prosper, Jesse V.
On Wed, Sep 11, 2013 at 11:10:07AM -0600, Jesse Victors wrote:
Do I have to maintain an uptime of ~70 days to see fully utilization then? This relay is on a personal computer with a static IP, so it isn't on a dedicated server or anything like that. Usually my uptime is around several weeks though. I'll be on this same IP for months though.
No, it's ok to be up only some of the time. Most of the question is what fraction of total time you're up ("weighted fractional uptime"), not how long you're up before restarts ("mean time between failure", which is used for calculating the Stable flag). Or said another way, the more time you miss, the fewer clients will pick you -- it has to do with whether you are in the set of possible choices at the moment when they decide to get rid of a current guard from their list and replace it. --Roger
On 09/11/2013 07:20 PM, Jesse Victors wrote:
Hello everyone, newcomer here.
I'm behind very fast connection (11.5 MB/sec down, 7.5 MB/sec up) and I thought that the Tor network could benefit from my connection, especially since it's apparently been under high load recently. Per the latest blog posts, I downloaded the beta TBB and configured it as a relay under Linux. It's been up for almost two days now, yet it's still being utilized at a very, very small fraction of it's potential. In the network map, I see that my relay has an advertised speed which is again much slower than it actually can be. To my knowledge, a web server can be put under full load right away, and distributing computing projects use the most of your computer right off the bat. Why doesn't Tor run computational and/or bandwidth tests and advertise my relay at a much more actual speed? I don't see why a fast relay has to start at the very bottom of the barrel to begin with. I can see the logic in slowly ramping up connections in testing for stability and scalability, but the progress is slow because it's continually fighting against that existing average. I can set my requested average and burst bandwidth, but AFAIK I can't affect my advertised bandwidth. Why are things this way, and is there anything I can do to speed up the utilization growth?
Thanks in advance for any help you can provide, Jesse V.
On 09/11/2013 12:43 PM, Roger Dingledine wrote:> Many people set up new fast relays and then wonder why their bandwidth
is not fully loaded instantly. In this post I'll walk you through the lifecycle of a new fast non-exit relay, since Tor's bandwidth estimation and load balancing has gotten much more complicated in recent years.
https://blog.torproject.org/blog/lifecycle-of-a-new-relay
--Roger
I think the best thing to do is be patient. :)
participants (4)
-
irregulator＠riseup.net -
Jeroen Massar -
Jesse Victors -
Roger Dingledine