On 08.12.2016 23:39, Alec Muffett wrote: > For general interest (perhaps to Juha, especially?): > > * I am building a 6-node, 24-core cluster, specifically to run an > Onion-traffic-serving experiment upon. > > * It's running a Debian variant - so the results/learnings should be > generally applicable to Linux - on Raspberry Pi hardware. Total cost ~ £400. > > * It's a casual-fun-research thing (rather than a big academic exercise) > and the current goal is to serve half-a-gigabit from a "single" onion > address, across ~30 daemons. > > * Details and diary are at: > https://github.com/alecmuffett/drafts-not-complete-not-tested-do-not-use/tree/master/onion-cluster-ideas > > Hopefully we should have results by end of January. :-) > Very interesting experiment! Keep us up-to-date please :)
I have no idea what will be the bottleneck of your cluster but it might be something very surprising. For instance, years ago I had some exit nodes and it took me several days to figure out what was the bottleneck of the traffic. There were plenty of CPU, RAM and bandwidth available but the entropy level of the VM was close to zero. Tip: follow your cat /proc/sys/kernel/random/entropy_avail and maybe apt-get install haveged. Good luck! -Juha -- tor-talk mailing list - [email protected] To unsubscribe or change other settings go to https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk
