Hi Wim,

 

First off I want to say that I greatly appreciate your willingness to engage
with and participate in the community here.  That alone says something very
strong to me in favor of continuing to advocate for Sun Ray and reduces my
worries about the future of the product in the education industry.  That
doesn't mean I don't have concerns though, and those I've addressed below.
I apologize in advance for the length, but I hope that it is justified in
the volume of information I would like to convey (which I haven't really
limited to just "Linux support"). 

 

To preface what I'm going to suggest, I should say that despite my e-mail
address being @tjhsst.edu, I actually speak on behalf of 3 groups with an
interest in Sun Ray.  The common ground is that they are all based out of
education/research institutions.  The groups:

 

1. Thomas Jefferson High School for Sci/Tech (TJHSST) student systems
administrators, responsible for a 150+ DTU deployment with concurrent usage
of about 45 users on average.  30 DTUs are deployed in an AP Computer
Science lab (students programming in Java).  7 are currently deployed as
kiosks (but standard login, not Sun Ray Kiosk mode) in common areas, with
plans for more, although some of that may need to wait until the building is
renovated in a couple years.  Most of the rest are deployed in our science
and technology senior research labs, as well as science department
classrooms.  We primarily run Solaris on x86 for Sun Ray, with one M4000
SPARC box for our parallel computing classes (non-Sun Ray) and for
applications that don't run on Solaris x86 (e.g. MATLAB).  We also have some
SPARC workstations for VirtualGL/Sun Ray, but we have not yet found an edu
application for that, and Sun dropped support for the VirtualGL project, so
I suspect Sun Ray support for VirtualGL will no longer be developed unless
Oracle picks the project back up.  In testing, we found that to be a pretty
good product for delivering 3D-accelerated OpenGL to a DTU (you can take
this as an indirect suggestion if you like).  Beyond Java programming, use
is primarily general desktop productivity.  I am an alumnus of TJHSST, where
I was a student admin, and currently I provide extended support for the Sun
infrastructure that I architected while I was still a student.  We have a
support contract until 2012, but it is unknown if we will be financially
able to renew due to predicted budget shortfalls and Oracle discontinuing
lower support tiers and support discounts for education.

 

2. Laboratory for Computational Science and Robotics (LCSR)
research/engineering group at Johns Hopkins University, where I am currently
a student.  There is a ~30 DTU deployment on Ubuntu.  We haven't settled on
a version yet because of our desire to use NFSv4/krb5, but we've found a
number of bugs in our NAS and in the Linux implementation of NFSv4/krb5.  We
will be trying 10.04 next.  This is primarily to support access to large
computational resources.  I believe C/C++ developers and MATLAB users are
the primary audience for Sun Ray here.  No support contract.

 

3. IT-SIG (IT Sustainability Initiative Group) at JHU.  We are investigating
and evaluating thin clients for deployment into various IT spaces at JHU.
JHU has a very departmental IT organization, with only core infrastructure
centrally managed.  The environments that result from our efforts will most
likely be in support of Windows or a mixed environment, so either a terminal
server setup or VDI will be used if we go with Sun Ray/Windows.  I believe
most other vendors' offerings are Windows-native already.  Support contracts
highly probable for most resulting deployments.

 

List of things I'd like to see, roughly in order of important, most
important at the top:

 

Pricing: {All} Since OEL is free but pay for support, I would love to see
Solaris go back to that model, but that's probably not your realm of
management.  It would be great if SRSS had edu discounts for hardware,
licenses, and support again.  I would also like to see a return of the
concurrent user license (see thread here
http://www.mail-archive.com/[email protected]/msg15347.html,
although I do now realize there is a situation in which per-device could be
more economical than per-concurrent-session, and that would be with numerous
smart card users that are disconnected.  I still think customers should have
a choice though.).  Even better would be taking SRSS to free to use but pay
for support.

 

To expand on what was already said about video handling on Linux/UNIX:

{TJHSST} Projector stations in classrooms are perfect candidates for Sun
Rays because they often sit for long periods unused, but they must be able
to play fullscreen video, especially YouTube.  Playing Flash videos in
MPlayer is already much better than using the Adobe Flash plugin because of
XVideo enhancements, but is a big hassle and most users don't know they can
do this, let alone how.  I don't know what can be done for Flash, but I
suspect there is something.  Non-Flash video (which also won't do fullscreen
today) may be easier to improve, see
http://www.filibeto.org/pipermail/sunray-users/2010-February/014987.html
(Sidenote: VA-API was added to the recent VLC 1.1 release.)

{ITSIG} Other vendors have a leg up because they have better video handling,
even on Windows to some extent.  While those thin clients may be less
"thin", I think "thin" is not as strong a selling point as better video.

 

USB: {All} It would be nice if some form of Daniel Cifuentes' usbdrived was
bundled into SRSS in the future.  Apart from that, USB performance in
general could use significant improvement as USB drives are now ubiquitous
in nearly every environment.  Students expect to be able to use them at
reasonable speeds, while throughput today is sub-USB 1.1.  I realize that
since DTUs do full USB abstraction over the network and the inherent need of
USB for zero latency makes this difficult to do, but maybe it would be worth
making it less "abstract" in order to imrpove throughput.

 

Application support: {All} Work with vendors to keep or add Sun Ray or
Solaris support on their apps.  MATLAB dropped Solaris/SPARC recently and
hasn't added Solaris/x86.  {TJ} TestNav (Pearson), used for state
standardized testing, is a Java program packed for Win/Mac and runs fine
under WINE but is not officially supported on *nix or recommended for any
type of thin client deployment.  {All} SmartBoards were once supported on
Solaris and Sun Ray but haven't been for many versions.  NI LabView dropped
Solaris support a few years ago, and it's unclear if Sun Rays were ever
supported on Solaris or Linux, even unofficially

 

Ubuntu: {LCSR} To echo and add on to what others have already said, Ubuntu
is user-friendly, easy-to-adminster, and has a very extensive package
repository.  I would argue that Debian and Ubuntu have the widest package
selection out of all commonly used Linux distros.  To tag on to that, the
widest package selection includes a lot of developer packages that our users
need, and we would need to invest significant time into compiling/packaging
these tools and libs on other distros.  Also, Ubuntu tends to be more up to
date on app versions compared to enterprise linux distros, but of course,
that also makes it harder for Oracle developers to keep up with kernel
releases/interfaces.  To be honest, I think I would largely be happy with
Ubuntu LTS being officially supported and other versions on a best-effort
but not necessarily guaranteed basis.

 

2FS/3+ Xinerama: {All} Xinerama between the two screens.  There is
fakexinerama for Linux, but it doesn't work with the Sun Ray X server as
Xinerama seems to be statically linked.

 

JDS: {TJ} We have used Solaris 10 primarily because it was the best option
at the time we were deploying new Sun Ray servers and because it offers JDS.
JDS enabled us to deploy DTUs to users without any additional training,
since we are primarily a Windows shop (but don't have TS CALs and don't
currently want to do the VDI solutions that are offered by vendors).  Had it
been vanilla GNOME, KDE, or something else, we either would have to do
significant customization work, perform significant training of users, or
meet user resistance to Sun Ray deployment.  While this is indirect, the
availability of the JDS/one-panel layout on Solaris and/or Linux is key for
the longevity and continued expansion of Sun Ray here.

 

Fluid VDI: {All} VDI is too inflexible in my opinion. Prior to the release
of VDI 3, I imagined that the best possible solution was actually a cross
between what is now VDI 3 and what was the old Virtual Desktop Connector.
There would be a group of servers, each server running SRSS and VirtualBox
(i.e. single-tier solution).  Non-kiosk mode would launch native sessions
and kiosk mode could be configured for any number of options, one of which
to operate in "VDI mode" and connect to a VirtualBox VM.  This would offer
the ability to operate a mixed-OS environment in the most flexible and
resource-efficient way since the entire group of servers could be sized for
a range of the total number of users instead of having to size for both max
native sessions in one server group and max VDI VMs in another group.  In
addition, native sessions are the most efficient and feature-rich way to
have a *nix session.  It is possible to do this today by switching around
the kiosk policy in VDI 3 on a per-token basis, but that is unsupported and
tends to make the whole setup unstable since SRSS is not in the state VDI
assumes it to be in.  Maybe this is something that could be done in the
community if Oracle is not interested.  Ideally it would not require any
additional licenses on top of the standard SRSS one, as was the case with
the old VDC.

 

Mobile DTUs: {TJ, IT-SIG} I think this was mentioned in some form earlier in
the thread, but a Sun Ray in a mobile form factor that is more than
competitive with laptops would be appreciated.  Currently the best mobile
Sun Rays seem to only have 3-4 hours battery life and 802.11g wireless.  I
would think with the low power processor, the battery life could be much
better, especially as netbooks can get really long battery life.  Perhaps
this is more a result of the power drawn by the wireless though.  They also
seem to be just as heavy as a standard laptop of the same screen size.  It
would be great if improvements could be made since laptops are becoming
increasingly used these days.

 

Audio: {TJ} Currently only one device can talk to a DTU's audio devices.  On
Linux this isn't so much a problem because of soundservers, but on Solaris
10, the lack of a software mixer interface is somewhat annoying since I need
to launch a utaudio daemon to mix multiple applications in hardware.  I was
told this RFE was a nogo when I filed it with Sun support because of
OpenSolaris/OSS being the way forward, but since that project currently
appears stalled.. 

 

Other feedback:

 

I have used VirtualBox with a Solaris+SRSS development/testing/SRSS-to-go
setup on my laptop, but we currently don't run virtualized SRSS in
production.  We are considering it if it can be done with very minimal
overhead and no additional cost as it could allow us to migrate SRS's around
while performing maintenance on physical hardware, or to shut down extra
servers to conserve power, etc.  However, I think preinstall images would
only really be more useful for VDI setups since users don't see the UNIX OS
as much in that case.  For instance, I run SRSS on Solaris and Ubuntu, but
not any EL, so for a preconfigured image to be useful for me, it would need
to come on one of those two platforms.  I would not, however, use VDI to
access a UNIX backend.  It seems to be too much overhead (1 OS instance per
user) and complexity compared to native logins, unless I were supporting an
IT class where every student needed root on their own VM.

 

Someone suggested free giveaways to computer clubs, and I would give that a
hearty +1.  I say that as both a beneficiary of Sun's former edu grant
programs (without which I may never have even heard of Sun Ray) and as
someone that agrees that hobbyists at educational institutions are where
technologies need to get exposure to last long-term (I apply this
perspective to Solaris here:
http://wyang0.blogspot.com/2010/03/long-term-investment-or-solaris-is.html).


 

Thanks for reading and sorry again for the very lengthy e-mail!  I hope I
didn't discourage you from reading.

 

William Yang

_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users

Reply via email to