Free estimator software for linux

BPF Rate Estimator estimates the traffic rate made by a BPF Expression.

Installation:

./configure [options]
make
make install

Usage: bpfrate [-i interface] bpf_expr
Calculate trafic rate made by an "bpf expression"
-i interface - interface to listen on (default eth0)
bpf_expr - bpf expression to apply - see tcpdump and pcap (default none)
-h - display this help

Mantissa is a collection of various mathematical tools aimed towards for simulation.
Mantissa contains a collection of algorithms, among which:
a small set of linear algebra classes
a least squares estimator
some curve fitting classes
several ordinary differentials equations integrators, either with fixed steps or adaptive stepsize control (see below)
vectors and rotations in a three dimensional space
algebra-related classes like rational and double polynomials
various orthogonal polynomials:
Chebyshev
Hermite
Laguerre
Legendre
some random numbers and vectors generation classes:
Robert M. Ziff four tap shift register (contributed by Bill Maier)
Makoto Matsumoto and Takuji Nishimura Mersenne twister
generators for vectors with correlated components
some basic (min, max, mean, standard deviation) statistical analysis classes
some optimization algorithms using direct search methods:
the Nelder-Mead simplex method
Virginia Torczons multi-directional method
Enhancements:
- For many basic objects provided by Mantissa like Vector3D, Rotation, and the various Polynomial classes, instances are now guaranteed to be immutable.
- This greatly simplifies safe sharing of instances without forcing users to either put severe restrictions on their use of Mantissa classes or make numerous copies just to make sure everything is safe.
- Since the change is a semantic change on the contract of the classes, this version introduces some incompatibilities with respect to previous ones.
- Upgrading to this version is not difficult, though.

The Wonder Shaper is a very special network shaper script with a lot of features. Works on Linux 2.4 & higher.

Goals

I attempted to create the holy grail:

* Maintain low latency for interfactive traffic at all times.

This means that downloading or uploading files should not disturb SSH or even telnet. These are the most important things, even 200ms latency is sluggish to work over.

* Allow surfing at reasonable speeds while up or downloading

Even though http is bulk traffic, other traffic should not drown it out too much.

* Make sure uploads dont harm downloads, and the other way around

This is a much observed phenomenon where upstream traffic simply destroys download speed. It turns out that all this is possible, at the cost of a tiny bit of bandwidth. The reason that uploads, downloads and ssh hurt eachother is the presence of large queues in many domestic access devices like cable or DSL modems.

Why it doesnt work well by default

ISPs know that they are benchmarked solely on how fast people can download. Besides available bandwidth, download speed is influenced heavily by packet loss, which seriously hampers TCP/IP performance. Large queues can help prevent packetloss, and speed up downloads. So ISPs configure large queues.

These large queues however damage interactivity. A keystroke must first travel the upstream queue, which may be seconds (!) long and go to your remote host. It is then displayed, which leads to a packet coming back, which must then traverse the downstream queue, located at your ISP, before it appears on your screen.

This HOWTO teaches you how to mangle and process the queue in many ways, but sadly, not all queues are accessible to us. The queue over at the ISP is completely off-limits, whereas the upstream queue probably lives inside your cable modem or DSL device. You may or may not be able to configure it. Most probably not.

So, what next? As we cant control either of those queues, they must be eliminated, and moved to your Linux router. Luckily this is possible.

Limit upload speed somewhat

By limiting our upload speed to slightly less than the truly available rate, no queues are built up in our modem. The queue is now moved to Linux.

Limit download speed

This is slightly trickier as we cant really influence how fast the internet ships us data. We can however drop packets that are coming in too fast, which causes TCP/IP to slow down to just the rate we want. Because we dont want to drop traffic unnecessarily, we configure a burst size we allow at higher speed.

Now, once we have done this, we have eliminated the downstream queue totally (except for short bursts), and gain the ability to manage the upstream queue with all the power Linux offers.

Let interactive traffic skip the queue

What remains to be done is to make sure interactive traffic jumps to the front of the upstream queue. To make sure that uploads dont hurt downloads, we also move ACK packets to the front of the queue. This is what normally causes the huge slowdown observed when generating bulk traffic both ways. The ACKnowledgements for downstream traffic must compete with upstream traffic, and get delayed in the process.

We also move other small packets to the front of the queue - this helps operating systems which do not set TOS bits, like everything from Microsoft.

Allow the user to specify low priority traffic (new in 1.1!)

Sometimes you may notice low priority OUTGOING traffic slowing down important traffic. In that case, the following options may help you:

NOPRIOHOSTSRC
Set this to hosts or netmasks in your network that should have low priority

NOPRIOHOSTDST
Set this to hosts or netmasks on the internet that should have low priority

NOPRIOPORTSRC
Set this to source ports that should have low priority. If you have an unimportant webserver on your traffic, set this to 80

NOPRIOPORTDST
Set this to destination ports that should have low priority.

See the start of wshaper and wshaper.htb

Results

If we do all this we get the following measurements using an excellent ADSL connection from xs4all in the Netherlands:

Baseline latency:
round-trip min/avg/max = 14.4/17.1/21.7 ms

Without traffic conditioner, while downloading:
round-trip min/avg/max = 560.9/573.6/586.4 ms

Without traffic conditioner, while uploading:
round-trip min/avg/max = 2041.4/2332.1/2427.6 ms

With conditioner, during 220kbit/s upload:
round-trip min/avg/max = 15.7/51.8/79.9 ms

With conditioner, during 850kbit/s download:
round-trip min/avg/max = 20.4/46.9/74.0 ms

When uploading, downloads proceed at ~80% of the available speed. Uploads at around 90%. Latency then jumps to 850 ms, still figuring out why.

What you can expect from this script depends a lot on your actual uplink speed. When uploading at full speed, there will always be a single packet ahead of your keystroke. That is the lower limit to the latency you can achieve - divide your MTU by your upstream speed to calculate. Typical values will be somewhat higher than that. Lower your MTU for better effects!

A small table:

Uplink speed | Expected latency due to upload
--------------------------------------------------
32 | 234ms
64 | 117ms
128 | 58ms
256 | 29ms

So to calculate your effective latency, take a baseline measurement (ping on an unloaded link), and look up the number in the table, and add it. That is about the best you can expect. This number comes from a calculation that assumes that your upstream keystroke will have at most half a full sized packet ahead of it.

This boils down to:

mtu * 0.5 * 10
-------------- + baseline_latency
kbit

The factor 10 is not quite correct but works well in practice.

Your kernel

If you run a recent distribution, everything should be ok. You need 2.4 with QoS options turned on.

If you compile your own kernel, it must have some options enabled. Most notably, in the Networking Options menu, QoS and/or Fair Queueing, turn at least CBQ, PRIO, SFQ, Ingress, Traffic Policing, QoS support, Rate Estimator, QoS classifier, U32 classifier, fwmark classifier.

In practice, I (and most distributions) just turn on everything.

The scripts

The script comes in two versions, one which works on standard kernels and is implemented using CBQ. The other one uses the excellent HTB qdisc which is not in the default kernel. The CBQ version is more tested than the HTB one!

See wshaper and wshaper.htb.

Tuning

These scripts need to know the real rate of your ISP connection. This is hard to determine upfront as different ISPs use different kinds of bits it appears. People report success using the following technique:

Estimate both your upstream and downstream at half the rate your ISP specifies. Now verify if the script is functioning - check interactivity while uploading and while downloading. This should deliver the latency as calculated above. If not, check if the script executed without errors.

Now slowly increase the upstream & downstream numbers in the script until the latency comes back. This way you can find optimum values for your connection. If you are happy, please report to me so I can make a list of numbers that work well. Please let me know which ISP you use and the name of your subscription, and its reputed specifications, so I can list you here and save others the trouble.

Installation

If you dial in, you can copy the script to /etc/ppp/ip-up.d and it will be run at each connect.

If you want to remove the shaper from an interface, run wshaper stop. To see status information, run wshaper status.

KNOWN PROBLEMS

If you get errors, add an -x to the first line, as follows:

#!/bin/bash -x

And retry. This will show you which line gives an error. Before contacting me, make sure that you are running a recent version of iproute!

Recent versions can be found at your Linux distributor, or if you prefer compiling, here:
ftp://ftp.inr.ac.ru/ip-routing/iproute2-current.tar.gz

Gretl is a cross-platform software package for econometric analysis, written in the C programming language. It is is free, open-source software.
You may redistribute it and/or modify it under the terms of the GNU General Public License (GPL) as published by the Free Software Foundation.
Main features:
- Easy intuitive interface (now in French, Italian, Spanish, Polish and German as well as English)
- A wide variety of least-squares based estimators, including two-stage least squares and nonlinear least squares
- Single commands to launch things like augmented Dickey-Fuller test, Chow test for structural stability, Vector Autoregressions, ARMA estimation
- Output models as LaTeX files, in tabular or equation format
- Integrated scripting language: enter commands either via the gui or via script
- Command loop structure for Monte Carlo simulations and iterative estimation procedures
- GUI controller for fine-tuning Gnuplot graphs
- Link to GNU R for further data analysis