[tor-relays] "Relay info kit" for Tor exits at universities

[Scott Bennett]

     On Sat, 11 Aug 2012 04:08:57 -0400 Roger Dingledine <arma at mit.edu>
wrote:
>I've spent the week talking to my contacts at US universities, to get
>them to spin up fast exits. Currently the fast exits in North American
>universities are:
>
>- University of Waterloo (Ian Goldberg)
>- Boston University (Leo Reyzin)
>
>We're now on track to add:
>- UPenn (Matt Blaze)
>- UMichigan (Alex Halderman)
>- CMU (Nicolas Christin)
>- Georgia Tech (Dave Dagon)
>
>and I have professors from George Mason, Illinois, UNM, UMN, UConn,
>UW, and others looking into it.
>
>Wendy and I are talking to some lawyers to try to write up a short
>(several paragraph) document targeted toward the university's general
>counsel, for preemptive use by the computer science professors who plan
>to run the Tor exit.
>
>What else should go in a "so you want to run a big exit" info kit?
>
>- Pointers to the legal-faq (and dmca template) and abuse-faq.
>
>- Pointers to Mike's blog entry:
>https://blog.torproject.org/blog/tips-running-exit-node-minimal-harassment
>and my old Tor-at-universities wiki page:
>https://trac.torproject.org/projects/tor/wiki/doc/TorGuideUniversities
>
>- Is there some document suggesting how to SWIP your address, and
>explaining the importance of having your abuse mails go to someplace
>other than your general university abuse team? It's touched on in several
>places but we should make it even clearer.
>
>- What are the good answers now to "what hardware should I use, and how
>should I configure it?" I've been telling people they'll be happiest with
>Debian, and that something 64-bit and/or with AESNI support will be best.

     Fast relays use lots of memory, which vastly worsens the effects of
the "small TLB" problem inherent in all modern CPUs, so running a fast relay
on a FreeBSD kernel with superpages support enabled is likely to reduce
considerably the CPU load for the traffic served, which is more or less
equivalent to increasing the traffic capacity for any given CPU speed.
     Superpages support has been available since 7.2 for i386 and amd64
platforms, but will likely be available on other platforms at some point.
Support on another platform was supposed to appear in 9.0, IIRC, but I don't
remember which (ia64?).  I'm not sure what will be in later releases, but my
understanding is that the kernel development team planned to add superpages
support eventually to the rest of the FreeBSD-supported hardware platforms
that are capable of it (not, for example, powerpc because that hardware does
not support the multiple page sizes involved).
     Rather than risk triggering OS religious wars here, it might be better
to limit OS recommendations to merely which to avoid because they have
limitations likely to prevent successful operation of a fast exit node
(e.g., non-server versions of Windows).  The administrators of fast nodes
already know what their current setups are and the current knowledge bases
of their staffs are.
>
>- [remainder deleted  -SB]


                                  Scott Bennett, Comm. ASMELG, CFIAG
**********************************************************************
* Internet:       bennett at cs.niu.edu                              *
*--------------------------------------------------------------------*
* "A well regulated and disciplined militia, is at all times a good  *
* objection to the introduction of that bane of all free governments *
* -- a standing army."                                               *
*    -- Gov. John Hancock, New York Journal, 28 January 1790         *
**********************************************************************