Hi there, I'm the operator of the following two exit nodes: https://metrics.torproject.org/rs.html#search/family:94C268630BEDCB64E7F8881... My CPU is an Intel C2750 (8 cores total), which supports hardware accelerated AES yet a single process still maxes out at only ~150mbit/s. To get the most out of the machine and my single IPv4 address, I simply ran another relay which works fine, and now I'm averaging at around ~315mbit/s. Is there anything else I can tweak to improve throughput without having to order more IP addresses? I'm running Arch with stock 5.2.9 linux kernel. That being said, true multi threading would be much appreciated and would help people like me a lot :-) Regards, Jochen
A few years ago I had some Intel Atoms with Scaleway. Same problem. Its the best a Atom can do :/ niftybunny
On 26. Aug 2019, at 11:06, Jochen <jochen@cock.lu> wrote:
Hi there, I'm the operator of the following two exit nodes:
https://metrics.torproject.org/rs.html#search/family:94C268630BEDCB64E7F8881...
My CPU is an Intel C2750 (8 cores total), which supports hardware accelerated AES yet a single process still maxes out at only ~150mbit/s.
To get the most out of the machine and my single IPv4 address, I simply ran another relay which works fine, and now I'm averaging at around ~315mbit/s.
Is there anything else I can tweak to improve throughput without having to order more IP addresses? I'm running Arch with stock 5.2.9 linux kernel.
That being said, true multi threading would be much appreciated and would help people like me a lot :-)
Regards,
Jochen
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
Hi all, Jochen sent this reply to the list, with a 4 MB attachment, which is a bit big. I don't think it contains anything sensitive, but I'll check before forwarding it to anyone else. I've discarded the message and attachment from the moderation queue. Here's the message:
Hi,
I have attached a profile of the main thread, over 30 seconds, to this e-mail.
Let me know if you need anything else.
Regards,
Jochen
T -- teor ----------------------------------------------------------------------
Right now, Tor is not multithreaded. This is a well-documented issue. You have a two options: * Get multiple IPv4 addresses and explain Tor is not multithreaded as the justification * Consider getting one with a good single-core performance as opposed to more cores (I don't know if Scaleway will give you this, so consider switching if you have to) * If your CPU supports Intel Quick Assist, consider using this to accelerate crypto/SSL. If supported, you may not have to get more IPv4 addresses or change servers (I have not tried this, though). -Neel === https://www.neelc.org/ On 2019-08-26 05:06, Jochen wrote:
Hi there, I'm the operator of the following two exit nodes:
https://metrics.torproject.org/rs.html#search/family:94C268630BEDCB64E7F8881...
My CPU is an Intel C2750 (8 cores total), which supports hardware accelerated AES yet a single process still maxes out at only ~150mbit/s.
To get the most out of the machine and my single IPv4 address, I simply ran another relay which works fine, and now I'm averaging at around ~315mbit/s.
Is there anything else I can tweak to improve throughput without having to order more IP addresses? I'm running Arch with stock 5.2.9 linux kernel.
That being said, true multi threading would be much appreciated and would help people like me a lot :-)
Regards,
Jochen
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
Hi,
On 26 Aug 2019, at 19:06, Jochen <jochen@cock.lu> wrote:
Hi there, I'm the operator of the following two exit nodes:
https://metrics.torproject.org/rs.html#search/family:94C268630BEDCB64E7F8881...
My CPU is an Intel C2750 (8 cores total), which supports hardware accelerated AES yet a single process still maxes out at only ~150mbit/s.
To get the most out of the machine and my single IPv4 address, I simply ran another relay which works fine, and now I'm averaging at around ~315mbit/s.
Is there anything else I can tweak to improve throughput without having to order more IP addresses? I'm running Arch with stock 5.2.9 linux kernel.
That being said, true multi threading would be much appreciated and would help people like me a lot :-)
Which code is your relay spending most of its (main thread) time on? If you can send us a profile, we will know where to focus our effort. T
participants (4)
-
Jochen -
Neel Chauhan -
niftybunny -
teor