Re: How bandwidth requirement could be reduced when using thin clients?
Knut Yrvin wrote:
Great. I'll quote from the documentation by Dr A V Le Blanc at the
Manchester Computing University of Manchester. I'we read the documents
from the link. It's some difference between Blancs recomendations, and
ours. We want to use PXE-boot or etherbooth with no local disk on the
thin client to reduce the point of failure, and exclude the need for a
bootloader at the local machine. Our mandate is not a dual system
solution. It's Skolelinux only.
Only Linux would be nice to have, but A) The teachers would lynch me and
B) Despite they develpment work we carried out on WINE for the dual boot
, theres still far to many national curriculum approved packages that
only work within windows. Even the schools webbased management system
requires IE, despite the fixes it needs to work with other browsers
would take less that a weeks work. It'll be at least 3 years before we
can defenestrate the school.
Network booting creates its own set of problems of course, last year we
deployed a WAN connecting all the schools in manchester to supply them
with internet access. Using a 10Mb link and a linux based cache/proxy +
filter in every school we were utlilising even at peak times about 20%
total bandwidth and started talking about a centralised LDAP
authentication server. At which point the backbone providers main
exchange caught fire crippling 3/4 of the network, and pretty much
stopped that idea in its tracks.
If you face a similar situation, then instead of just crippling internet
access, the computers themselves would become useless, unless of course
there was a local server. By having even a small amount of local drive
space, say a 16Mb CF card you can provide further redundancy.
It's also a requirement to tackle 2-8 Mbps bandwidth between the schools
and a central server-farm. In my experience an university often has at
least 10 or 100 Mbps between buildings. It's not posible to increase
the bandwidth to 10 or 100 Mbps in the near future.
IIRC When I tried to break Owens test room, it was 30 machines on 10Mb's
and they all handled opening open office at the same time without
difficulty. I can provide you with Owens email if its not on the
presentation and your interested in harder numbers.
The last and least requirement is that the solution can't have any
licencing cost.
Thats purely a requirement at the university as its being deployed
across several public clusters in addition to science and engineering labs.
Regards,
Ben
--
On two occasions I have been asked [by members of Parliament],
'Pray, Mr. Babbage, if you put into the machine wrong figures, will the
right answers come out?' I am not able rightly to apprehend the kind of
confusion of ideas that could provoke such a question.
Charles Babbage
Reply to: