first draft of prometheus manual page

[svn/Prometheus-QoS/.git] / README

============================================================================

Prometheus QoS - steal fire from your ISP !
"fair-per-IP" quality of service utility
requires Linux kernel with HTB qdisc enabled
GNU+ Copyright(G)2007, Michael Polak (xChaos)
Credits: Credits: CZFree.Net, Netdave, aquarius
...and Martin Devera (.cz) for his HTB qdisc (of course)
...and Jakub Walczak (.pl) for providing feedback and patches
...and Ing. Jiri Engelthaler (.cz) for bugfixes and Asus WL500 port
...and Dial Telecom (slightly expensive ISP) for chance to test it

Feedback: xchaos(at)arachne.cz
Homepage: http://gpl.arachne.cz

============================================================================

QoS (or Quality-of-service) is IPv4 traffic shaper replacement for Internet
Service Providers (ISP). Dump your vintage hard-wired routers/shapers
(C|sco, etc.) in favour of powerful open source and free solution !

Prometheus QoS generates multiple nested HTB tc classes with various rate 
and ceil values, and implements optional daily traffic quotas and data
transfer statistics (as HTML). It is compatible with NAT, both asymetrical
and symetrical, yet still provides good two-way shaping and prioritizing,
both upload and download. Prometheus QoS allows both "hard shaping" 
(reducing HTB ceil value for aggressive downloaders) and "soft shaping" 
(keeping HTB ceil, but reducing HTB prio, probably optimal solution for 
normal users).

Prometheus iGW was written in C<<1, which means it compiles simply with 
GNU C Compiler, and doesn't require any external liberaries (except libc)
and huge interpreter packages (like Perl or Java) to run. However, it 
depends on HTB algorithm hardcoded in Linux kernel. It is currently being 
tested in real-world enviroment to provide QoS services on 30 Mbps internet
gateway and proxy being used by 2000+ PCs connected to gateway using
CZFree.Net broadband community network.

Advantages over more straightforward traffic control scripts include
HTB fine tuning features (rate and ceil magic), data transfer statistics,
optional data transfer quotas, full NAT (both symetric and one way)
compatibility and optinal sharing of bandwith by IPs in completely 
different subnets.

Performance and scaling - current release: 

we run Prometheus QoS on Celeron 2.8 Ghz serving around 600 individual
traffic classes (fine tuning is using five user-defined prometheus.conf 
keyword) and another 2000 IPs sharing bandwith with certain other IPs 
("sharing-" keyword). Prometheus QoS is especially strong tool if you want 
IP's from different subnets to share the same traffic class.

With 30 Mbps (each way) total capacity of line, Cisco Catalyst 2950 on 
ISP side and up to cca 6000 packets per seconds, we were running into some 
problems with overall system load. We moved QoS from Athlon 1700 XP to 
Celeron 2.8 Ghz, and kept all SNAT related stuff (see optinal-tools directory)
on Athlon 1.7 Ghz, which alowed for peak throughput up to 10000 pps.

Performance fine tuning - history:

With kernel version 2.4.20 and release 0.2 we started to experience problems
at cca 1500 packets/sec. However, with new iptables indexing feature
implemented in 0.3 release, system load seems to be approximately 
10 times lower. Same HW was later shaping 2000 packets/sec without problems,
and it looked like comparable relatively low-end system should be able to do
traffic shaping for at least 10000 packets/sec (well, if HotSaNIC was turned
off, of course <g>). With 0.6 release and dynamicaly calculated iptables
indexing scheme we made it up to 6000 packets/sec, and then we ran into some
performance-related problems, which may be related to the fact we are doing
SNAT of 1000+ individual IP addresses on the same machine which is doing 
also the QoS: something on the way seems to be limited to 34 Mbps HD
(half-duplex, sum of upload and download) no matter what we try. Our ISP 
claims the fault is not on his side, so our next step will be to separate
traffic shaping and massive SNAT (IP masquerading) and assign separate
PC-based router to do each task.

Maximum performance observed with prometheus 0.6 and hashtable optimization
of tables with individual SNAT targets was up to 9000 packets/sec at cca 40
Mbps half-duplex (more then 20 Mbps fyull-duplex). However, this required
massive optimization, including 
echo -n 65000 > /proc/sys/net/ipv4/ip_conntrack_max 
and
echo -n 21600 > /proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_established
and disabling of most userspace applications (like eg. hotsanic). At the
same time, router machine and system was accumulating wide set of various
performance related problems, which required us to reboot it at least
mohtly.

Note: Some time ago it seemed that maximum index of tc classes was restricted
to 10000 - but I haven't checked this again for quite a while.
Another note: All the echo stuff in previous paragraph can be also achieved by
adding following lines to /etc/sysctl.conf which is much cleaner way to do it:
sys.net.ipv4.ip_conntrack_max=65000
sys.net.ipv4.netfilter.ip_conntrack_tcp_timeout_established=21600

Future plans include also setting of individual daily limits on maximum
pps (packets per second) rates allocated to individual IP addresses (this
may be needed partly because of problems mentioned above).

============================================================================

Prometheus QoS is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as 
published by the Free Software Foundation; either version 2.1 of 
the License, or (at your option) any later version.

Prometheus QoS is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
General Public License for more details.

You should have received a copy of the GNU General Public License 
along with Prometheus QoS source code; if not, write to 
Michael Polak, Svojsikova 7, 169 00 Praha 6 Czech Republic