[cryptography] Introducing Bletchley

coderman coderman at gmail.com
Thu Aug 15 00:44:16 EDT 2013


On Wed, Aug 14, 2013 at 9:15 PM, Timothy D. Morgan
<tmorgan at vsecurity.com> wrote:
> ...
> Well, the goal of multithreading with POA is to overcome network latency.
> Almost all of the attacks I implement are online attacks against web apps.  POA
> doesn't require much CPU, it just requires sending lots of requests,

right; perhaps this isn't useful yet.

i was thinking of some particular very large socket and event handling
in a native module, however, on second thought this may not be needed
and where such optimization is useful it seems i can use an existing
module...


> ... Feel free to
> post an issue to the Google code site if you have any specific questions or just
> email me privately.  I can set up a mailing list once there's enough chatter.

please do setup a mailing list or google group if not too much
trouble. as for the feedback, not really a bug. here is some tuning i
was playing with for launching with many many threads/sockets on
linux. ymmv :)

--- NOTE: this config may require 16G or mem or more depending on
application and load.

# disable conntrack if possible
iptables -t raw -A notracking -j NOTRACK

# append to /etc/security/limits.conf
# launch new session, then raise ulimit s
*               soft    nofile          65535
*               hard    nofile         1048576

# echo for sysctl.conf; sysctl -p /etc/sysctl.conf
fs.file-max = 262140
net.core.somaxconn = 1048576
net.core.netdev_max_backlog = 2500
net.core.rmem_max = 16777216
net.core.wmem_max = 16777216
net.ipv4.tcp_mem = 4096 87380 16777216
net.ipv4.tcp_rmem = 4096 87380 16777216
net.ipv4.tcp_wmem = 4096 65536 16777216
net.ipv4.tcp_max_syn_backlog = 2500
net.ipv4.tcp_window_scaling = 1
net.ipv4.tcp_timestamps = 0
net.ipv4.tcp_sack = 0
net.ipv4.tcp_dsack = 0
net.ipv4.tcp_no_metrics_save = 1
net.ipv4.tcp_fin_timeout = 5


More information about the cryptography mailing list