Hi everyone! To start this blog I’m posting my accepted proposal on Google Summer of Code. It is useful for everyone who wants to know what are the multiseat’s problems and what I will do to solve them. From this day forward I will post weekly the updates of the project.
The Proposal:
Use upstream pieces to make a fully plug-and-play USB multiseat solution.
Name: Lucas Nascimento Ferreira
e-mail: lnf07@c3sl.ufpr.br
website: www.inf.ufpr.br/lnf07
Synopsis
Over the last years many multiseat solutions have appeared, but many are
closed projects and some are open projects that never became stable on any
distribution. Nowadays a lot of features are being added in Xorg and Linux,
and now is possible to make a clear and general solution that distributions
can easily include it.
One of these features is the correct VGA arbitration in Xorg 1.7 and
Linux Kernel 2.6.32 [0]. This allows to set up a multiseat using multiple X
servers on different graphical devices, and shows the interest of the X.org
community on this area. Another feature that is important to multiseat
solutions is the secondary video cards initialization [1], that was included
in Xorg 1.7 too. These features make possible to set up a multiseat with
multiple instances of X server, but a tool that is extremely useful to
manage sessions and seats is the ConsoleKit. All these pieces together and
correctly integrated make a clear multiseat solution.
Two common problems in multiseat environments are the association and
detection of the devices, one solution based on a USB hierarchy [2] makes
this environment more controlled and this turns the problem easier to be
solved. A USB hierarchy is a way of dispose the devices, a set of devices
(keyboard, mouse, ...) are attached in one USB HUB and this HUB is
associated to one user, making one seat. The Plugable HUBS [3] are more
useful to make a USB hierarchy because the video is also in the HUB, so the
seats are totally separated. This kind of solution has some problems,
because those USB HUBS must be fully plug-and-play, because user doesn't
want to manually reconfigure the devices every time a USB HUB is detached
from the machine. This problem can be solved with the udev[4] tool, that
provides a way to do a dynamic detection and association of the devices.
A USB Multiseat solution based on Plugable HUBS has another advantage, and
this is the main attraction to end users, the cost/benefit. The USB
terminals themselves are a significant cost savings over buying even
low-costs netbooks. The cost of maintaining PCs goes way down, as many
terminals share a single server. Plug and play nature also provides big
savings. This solution is perfect to "office user", in other words, those
users who only use browsers, messengers, text editors and other common
applications like these.
A hard work must be done to correctly integrate the new features + Linux
Kernel tools (udev) + ConsoleKit + display managers to make a fully
plug-and-play multiseat solution. What I am going to do is to study the
state of all these pieces nowadays and, using the existing source code [5]
of Plugable solution, refine and document it to make easily for
distributions to include it.
Why Multiseat ?
I would like to work in the "USB Multiseat Refinement" idea because I think
is very important that linux distributions have an easy way to configure a
multiseat environment. This is because in Brazil multiseats are becoming
very common on public schools and I have saw that they are having many
problems with it. Another big incentive to work with multiseats is because
this kind of solution decreases the cost of computing, so this is a good
tool to the digital inclusion. Three important points that justify the
advantages of multiseats are the energy-saving, CPU optimization and ease of
management. The first point is the most important, because with multiseats
is possible to increase the number of users in a lab just using a little
more of energy. This is because there is a large decrease of devices (CPUs,
hard disks, etc) by the total number of users. The second point is because
in a common "office environment" (with common computers) which people are
using only applications like browsers, messengers, text editors, a great part
of time the CPU is idle. This is because the user is reading a document,
seeing a website that is already loaded or something like these. So if in
this environment is changed the common computers to multiseats, the time
that the CPU is idle will be decreased, optimizing the use of CPU. The last
point occurs due to the decrease of computer amount, so it turns easier the
network management and the hardware maintenance.
Description
To create a fully plug-and-play USB multiseat solution is necessary to solve
two common problems of multiseats: detection and association of devices.
In a USB multiseat environment is necessary to detect every hub connected to
the machine. Actually we need to detect the devices, but they will be
plugged on the machine through the hub, for solve it will be created udev
rules. Every hub connected will be a seat, so is necessary to create some
udev attributes to label this set of devices with a common seat ID. With
this settings the detection problem is solved.
The association problem will be divided into two classes: basic devices and
other devices. What is considered basic devices are keyboard, mouse and
video. Other devices are storage/audio devices. To solve it we need to get
the devices that were detected and labeled through the udev rules/attributes
and write a ConsoleKit script to launch independent GDM/X session to each
set of devices (HUB). Before it, is necessary to check if in each hub are
connected the 3 basic devices (keyboard, mouse and video). Then is important
to create some InputClass rules to cause the primary X session to ignore
multiseat-assigned devices, and the appropriate seat to use them. Other
devices like pendrives, mp3 players which will be attached on the machine
when the sessions is already started, will also need to be detected and
associated correctly. Since each seat has its own HUB, each new device
attached into a hub must be associated and become available to that session
user. By creating some udev rules and X init scripts this can become
automatic.
Project Plan
Following the Google Summer of Code Timeline, these is my program schedule:
April 27 - May 24
Basically it is a study period, but detailed it is:
- Get some Display Link USB Hubs and set up a multiseat environment with
the source code that already exists.
- Study the existent source code to become expert in what already been
done.
- Understand how does udev works
- Study the state of Xorg today to know how to set up a good
configuration file and explore the new features.
- Study if there is some configuration or module that will need to set
up on kernel.
May 24 - June 7
- Define the udev rules to detect hubs/devices which should be
collectively treated as terminals
- Define the udev attributes to label the set of devices with a common
seat id
June 8 - June 30
At this period I will be more busy because I will be at the end of
semester and I will have a lot of exams.
- udev triggers for on-demand generation of the appropriate Xorg config
files, to allow seats to coexist with the primary display/devices.
- Implement ConsoleKit scripts to launch independent GDM/X sessions for
each USB terminal seat
June 1 - July 15
- Write documentation and make a lot of tests (performance,
plug-and-play tests, ...)
July 16
- Submit mid-term evaluation.
July 17 - July 29
- InputClass rules to cause the primary X session to ignore
multiseat-assigned devices, and the appropriate seat to use them
- udev rules and X init scripts to grant access to audio, storage, and
other devices to the person logged into the matching seat
July 30 - August 9
- Study the state of some of the principal distros today to assess the
best way to include a multiseat solution on it.
- Adapt the code to include the solution to an example distribution, like
Debian or Gentoo.
- Write documentation and make a lot of tests (performance,
plug-and-play tests, ...)
Personal Information
My name is Lucas Nascimento Ferreira, I'm a 20 years old student of
Computer Science at UFPR [6] (Federal University of Paraná). I live in
Curitiba, capital of Paraná state, Brazil. I have worked with multiseats
over the last 3 years in MDM project [7] at C3SL [8] (Center of
Scientific Computing and Free Software). In 2008 I wrote a considerable
amount of code [9] to MDM project and in 2009 I published a summary
[10, 11] (only in portuguese) about the performance of the X Window
System on the labs of the Informatics Department of UFPR. I have done
one presentation (only in portuguese) [12] about multiseats to students
of UFPR explaining the common problems and the MDM solution. From 2009
I started to work with NX server of NoMachine and Google's NeatX.
References
[0] http://bugs.freedesktop.org/show_bug.cgi?id=20817
[1] http://bugs.freedesktop.org/show_bug.cgi?id=20816
[2] http://www.ndiyo.org/systems/hubster/basics
[3] http://plugable.com/products/UD-160-A
[4] http://www.kernel.org/pub/linux/utils/kernel/hotplug/udev.html
[5] http://plugable.com/2009/11/16/setting-up-usb-multiseat-with-displaylink-on-linux-gdm-up-to-2-20
[6] http://www.ufpr.br
[7] http://wiki.c3sl.ufpr.br/multiseat
[8] http://www.c3sl.ufpr.br
[9] http://git.c3sl.ufpr.br/gitweb?p=multiseat/mdm.git;a=summary
[10] http://www.inf.ufpr.br/lnf07/en_us/publicacoes.html
[11] http://www.inf.ufpr.br/lnf07/arquivos/apresentacoes/apresentacao-17evinci.pdf
[12] http://www.inf.ufpr.br/lnf07/arquivos/apresentacoes/Multiterminal.pdf
0 Responses to “My GSOC 2010 Proposal: Use upstream pieces to make a fully plug-and-play USB multiseat solution.”