Thanks to the kind financial donation of @jonah from @triplebit, we've finally turned on our new SiFive HiFive P550 RISC-V Tor exit relay! this is our first ever non- x86-64 CPU.

https://metrics.torproject.org/rs.html#details/B8A3D6836E3DD183653814A399CD9FB93ED9D27D

The board: https://www.sifive.com/boards/hifive-premier-p550

If you want to help us run more hardware diversity to enhance the Tor network, please get in touch to work out a strategic donation!

Our lead relay engineer @alexhaydock has increased our stateless #Tor exit relay deployment to 96! (+1 because of the new #RISCV bare-metal node, +1 other we redeployed due to a silly spelling error). We're stress testing our three AMD Epyc 7402P servers that use #Proxmox.

Each one of the 96 Tor exit nodes are diskless Unified Kernel Images, 56MB in total size, using @alpinelinux's alpine-make-rootfs with an absolutely bare minimum number of packages. We'll be publishing more about our new architecture and configuration soon.

#AlpineLinux #privacy #anonymity #AntiCensorship #AccessToInformation #TorOps #TorOperators

up to [809.23 MiB/s = 6.79 Gbps] advertised bandwidth!

now up to [944.13 MiB/s = 7.92 Gbps] advertised bandwidth

and we broke into the top 20 exit providers, based on family, and consensus weight

hopefully as a US provider, we can get back into the top 10 again and lessen this Euro-centrism

We are now up to [1112.29 MiB/s = 9.33 Gbps] advertised bandwidth, and more importantly 0.4448% consensus weight.

At 15th, to overtake @flokinet, this might take some time. However, most of our relays currently are less than 2 weeks old, so after 14 days, it's possible we move up in the exit family ranking.

It's clear that adding more relays helps increase "advertised bandwidth" quite a bit. Our actual bandwidth utilization is less. We expect our eventual, peak advertised bandwidth will quickly go beyond our physical limitation of 10 Gbps.

For comparison, on 2023-March-13, our exit probability reached 3rd at 4.9118%, advertised bandwidth of [1907 MiB/s = 16.00 Gbps], but our peak throughput at that time was only 5.2 Gbps symmetrical
We dropped in Advertised Bandwidth this past week, and despite moving past the 2 week "new relay" period for most of our relays, we did not see an increase in overall traffic. We've fallen back down to 20th place in this ranking. Running relays in the US makes it hard to compete with the way the consensus is done.

# Evolving Our Tor Relay Security Architecture

https://blog.emeraldonion.org/evolving-our-tor-relay-security-architecture

A new blog post where @alexhaydock goes into some detail showcasing our minimalistic @alpinelinux Tor relays architecture, a threat model, and including a link to our now-public open source "Emerald Relays" orchestration framework.

The past 6 months have proven its success, and now we look forward to phase 2 (read the post!), which we need your help in funding. Emerald Onion is a U.S. 501(c)(3) tax-deductible nonprofit, so please consider donating before 31 December! https://emeraldonion.org/donate/

#Tor #TorOps #Privacy #AntiCensorship #Anonymity #Ansible #Proxmox #Terraform #AMD #Epyc #SEVSNP #NonProfit #GivingTuesday

Evolving Our Tor Relay Security Architecture

Fundraiser Emerald Onion needs your help! We are a U.S. tax-deductible 501\(c\)(3), and we are fundraising for new server hardware that s...

Emerald Onion