|
Home > Archive > Unix administration > January 2004 > Sorry, but any ideas fo a final-year uni project?
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Sorry, but any ideas fo a final-year uni project?
|
|
|
| Hi all,
I'm going to go into the final year of my Computer Networks &
Communications degree and need to come up with an idea for my
final-year project. Any suggestions or ideas will be gratefully
received.
I'm interested in Sun hardware and the Solaris OS and I've been
working with a lot of Sun gear as a sysadmin for my intern year. At
the moment I'm kind of thinking about producinga tool which reads the
tarred-up data output from running the explorer (SUNWexplo) package.
It's basically a package which collects a whole mass of diagnostic
data about a server and sends it to the engineers who can use it to
diagnose any problems the server had.
Another half-idea is to produce a tool which can map out an entire
network and then illustrate the state of the network on the screen
(i.e. hostnames, ping times, IP addresses etc). Trouble is, I'm pretty
sure this idea's been done a million times before and a million times
better.
A friend recommended I look at maybe investigating forensic databases
- I think the idea is basically that the database records every
relevant action that takes place on a server, like the logs, so that
it may be used reliably in any possible investigation. This opens up a
whole loads of issues like ensuring the database is 100% secure and
tamperproof.
Ideally I would like to do a project that does involve the
investigation of networks in some way, but I'm open to anything, and I
also would consider myself able to pick things up pretty easily after
a good bit of reading around the subject. I'm planning on learning a
lot of Java and maybe php in the next year too.
TIA
Cheers
Mark
| |
| Doug Freyburger 2004-01-23, 4:28 pm |
| Mark wrote:quote:
>
> I'm going to go into the final year of my Computer Networks &
> Communications degree and need to come up with an idea for my
> final-year project. Any suggestions or ideas will be gratefully
> received.
Do something usefull, needed, and not publicly available.
Read DNS plus NIS plus NIS+ plus hosts and build a database of
all known hosts. Compare with previous to map additions and
subtractions. Ping all known and previously known machines and
preport removed ones that still ping and new ones that don't.
The read the NIS or local netmasks and build a table of all
"local" IP numbers according the the IP population above. Have
some sort of exclude list in case folks add in remote machines.
If you have properly configued DNS reverse tables, use those.
Then scan for hosts that aren't in the tables. Send an e-mail
to report them. Arrange with the firewall folks so they expect
you to do ping scans.
Then read the NetBackup client configurations. Automagically add
new hosts into the default backup cycle. Automagically remove
decommissioned hosts from the backup cycle.
Hmmm, a backup system that doesn't have a client module to install
would be even better than NetBackup. Anyways, pick the backup
system that you site uses and autogenerate for it.
Autodiscovery and autobackup rules. I've done the above for sites
that use NIS, but not for sites that use DNS.
| |
| Doug Freyburger 2004-01-23, 4:28 pm |
| Mark wrote:quote:
>
> I'm going to go into the final year of my Computer Networks &
> Communications degree and need to come up with an idea for my
> final-year project. Any suggestions or ideas will be gratefully
> received.
Do something usefull, needed, and not publicly available.
Read DNS plus NIS plus NIS+ plus hosts and build a database of
all known hosts. Compare with previous to map additions and
subtractions. Ping all known and previously known machines and
preport removed ones that still ping and new ones that don't.
The read the NIS or local netmasks and build a table of all
"local" IP numbers according the the IP population above. Have
some sort of exclude list in case folks add in remote machines.
If you have properly configued DNS reverse tables, use those.
Then scan for hosts that aren't in the tables. Send an e-mail
to report them. Arrange with the firewall folks so they expect
you to do ping scans.
Then read the NetBackup client configurations. Automagically add
new hosts into the default backup cycle. Automagically remove
decommissioned hosts from the backup cycle.
Hmmm, a backup system that doesn't have a client module to install
would be even better than NetBackup. Anyways, pick the backup
system that you site uses and autogenerate for it.
Autodiscovery and autobackup rules. I've done the above for sites
that use NIS, but not for sites that use DNS.
| |
| Len Philpot 2004-01-23, 4:28 pm |
| In comp.unix.admin Mark <g_i_f_f_o@hotmail.com> wrote:quote:
> Hi all,
>
> I'm going to go into the final year of my Computer Networks &
> Communications degree and need to come up with an idea for my
> final-year project. Any suggestions or ideas will be gratefully
> received.
> I'm interested in Sun hardware and the Solaris OS and I've been
> working with a lot of Sun gear as a sysadmin for my intern year. At
> the moment I'm kind of thinking about producinga tool which reads the
> tarred-up data output from running the explorer (SUNWexplo) package.
> It's basically a package which collects a whole mass of diagnostic
> data about a server and sends it to the engineers who can use it to
> diagnose any problems the server had.
Here's a variation on that : Write a tool that will extract all the
recommended patches from a given host's Sentinel Report, recursively
including all their prereqs, and then work out the dependency order.
From that, generate a script of some sort (Expect, ftp, whatever) that
when run, will download all the patches and construct a cluster.
Some nice enhancements to the above would be :
* Option to include (or not) the "complete fix" patches
* Option to include/exclude patches of a given category, i.e., security,
X11, networking, etc., etc.
* Capability to compare a number of such patch groupings, build a common
cluster from those patches common to all systems and then construct
just diff clusters for each host.
* What else?
It's important to build the script (or response list, or whatever) ahead
of time instead of just downloading the patches on the fly. The reason
is that with a number of systems to patch, downloading the individual
patches might take some time. If a patch is obsoleted or superceded
during that time, your clusters for each system wouldn't stay in version
sync if you're just downloading patches based on the patch number only.
With a predefined list (including versions), at least you'd know that
when you went to get them.
Such a tool doesn't exist at Sun, at least not with all those
capabilities. There's PatchTool, patchcheck, patchdiag, etc., etc., but
none offer the Big Three : 1) Build the list, 2) Find dependencies & 3)
work out the application order. There would be some work to determine
the best way to interface with SunSolve, etc., but it would be great if
it could be done.
Such a tool would be gladly accepted at Sun sites where many systems are
to be patched on a regular basis.
--
-- Len Philpot ><> --
-- len@philpot.org http://philpot.org/ --
-- lphilpot01@yahoo.com (alternate email) --
| |
| Len Philpot 2004-01-23, 4:28 pm |
| In comp.unix.admin Mark <g_i_f_f_o@hotmail.com> wrote:quote:
> Hi all,
>
> I'm going to go into the final year of my Computer Networks &
> Communications degree and need to come up with an idea for my
> final-year project. Any suggestions or ideas will be gratefully
> received.
> I'm interested in Sun hardware and the Solaris OS and I've been
> working with a lot of Sun gear as a sysadmin for my intern year. At
> the moment I'm kind of thinking about producinga tool which reads the
> tarred-up data output from running the explorer (SUNWexplo) package.
> It's basically a package which collects a whole mass of diagnostic
> data about a server and sends it to the engineers who can use it to
> diagnose any problems the server had.
Here's a variation on that : Write a tool that will extract all the
recommended patches from a given host's Sentinel Report, recursively
including all their prereqs, and then work out the dependency order.
From that, generate a script of some sort (Expect, ftp, whatever) that
when run, will download all the patches and construct a cluster.
Some nice enhancements to the above would be :
* Option to include (or not) the "complete fix" patches
* Option to include/exclude patches of a given category, i.e., security,
X11, networking, etc., etc.
* Capability to compare a number of such patch groupings, build a common
cluster from those patches common to all systems and then construct
just diff clusters for each host.
* What else?
It's important to build the script (or response list, or whatever) ahead
of time instead of just downloading the patches on the fly. The reason
is that with a number of systems to patch, downloading the individual
patches might take some time. If a patch is obsoleted or superceded
during that time, your clusters for each system wouldn't stay in version
sync if you're just downloading patches based on the patch number only.
With a predefined list (including versions), at least you'd know that
when you went to get them.
Such a tool doesn't exist at Sun, at least not with all those
capabilities. There's PatchTool, patchcheck, patchdiag, etc., etc., but
none offer the Big Three : 1) Build the list, 2) Find dependencies & 3)
work out the application order. There would be some work to determine
the best way to interface with SunSolve, etc., but it would be great if
it could be done.
Such a tool would be gladly accepted at Sun sites where many systems are
to be patched on a regular basis.
--
-- Len Philpot ><> --
-- len@philpot.org http://philpot.org/ --
-- lphilpot01@yahoo.com (alternate email) --
| |
| Dr. David Kirkby 2004-01-23, 4:28 pm |
| Sorry your post has expired from my news server, so I don't know your
original question, so are going on the subject title.
How about a comparision of the many ways of getting better peformance
from multiple processing. There are good physical reasons you can't
just get better and better peformance from one cpu, so ultimiately
high-performance machines are going to have to use more than once
processing element. You might consider:
a) Multi-threaded prorgram, running multiple threads on different
cpus.
b) The approach of 'setiathome'
c) I once got involved in someone efforts to solve chess moves, but he
organised that badly, sending me the same data files more than once,
forgot to take into account I had multiple cpus .... etc. I gave up
with that.
c) OpenMP
d) MPICH.
e) Compilers that can automatically parallelise (ignore spelling) bits
of your program.
f) Intel's Hyperthreading.
g) and severl more.
You could consider peformance, ease of programming ...
That could easily make a PhD project, so you would need to be careful
to limit yourself to a small subset of all the possibilities.
Good luck.
PS, you could add in your dissertation you took suggestions from a
newsgroup. It shows initiative - something I look for when marking
undergrad/postgrad projects.
--
Dr. David Kirkby,
Senior Research Fellow,
Department of Medical Physics,
University college London,
11-20 Capper St, London, WC1E 6JA.
Tel: 020 7679 6408 Fax: 020 7679 6269
Internal telephone: ext 46408
e-mail davek@medphys.ucl.ac.uk
| |
| Dr. David Kirkby 2004-01-23, 4:28 pm |
| Sorry your post has expired from my news server, so I don't know your
original question, so are going on the subject title.
How about a comparision of the many ways of getting better peformance
from multiple processing. There are good physical reasons you can't
just get better and better peformance from one cpu, so ultimiately
high-performance machines are going to have to use more than once
processing element. You might consider:
a) Multi-threaded prorgram, running multiple threads on different
cpus.
b) The approach of 'setiathome'
c) I once got involved in someone efforts to solve chess moves, but he
organised that badly, sending me the same data files more than once,
forgot to take into account I had multiple cpus .... etc. I gave up
with that.
c) OpenMP
d) MPICH.
e) Compilers that can automatically parallelise (ignore spelling) bits
of your program.
f) Intel's Hyperthreading.
g) and severl more.
You could consider peformance, ease of programming ...
That could easily make a PhD project, so you would need to be careful
to limit yourself to a small subset of all the possibilities.
Good luck.
PS, you could add in your dissertation you took suggestions from a
newsgroup. It shows initiative - something I look for when marking
undergrad/postgrad projects.
--
Dr. David Kirkby,
Senior Research Fellow,
Department of Medical Physics,
University college London,
11-20 Capper St, London, WC1E 6JA.
Tel: 020 7679 6408 Fax: 020 7679 6269
Internal telephone: ext 46408
e-mail davek@medphys.ucl.ac.uk
| |
|
| g_i_f_f_o@hotmail.com (Mark) wrote in message news:<b97625a6.0307061323.3c71addd@posting.google.com>...
Hi everyone,
I'd just like to say a big thanks to everyone who took the time to
post a suggestion and it's definitely given me a few more ideas I
could pursue. I most like the patch tool idea, but after looking into
it I believe Sun have released a very similar application called Patch
Management System which is a module for the Sun Management Center. I
haven't had the time to look into it in too much detail but I
certainly will at some point in the future.
Once again, many thanks!
Cheers
Mark
| |
|
| g_i_f_f_o@hotmail.com (Mark) wrote in message news:<b97625a6.0307061323.3c71addd@posting.google.com>...
Hi everyone,
I'd just like to say a big thanks to everyone who took the time to
post a suggestion and it's definitely given me a few more ideas I
could pursue. I most like the patch tool idea, but after looking into
it I believe Sun have released a very similar application called Patch
Management System which is a module for the Sun Management Center. I
haven't had the time to look into it in too much detail but I
certainly will at some point in the future.
Once again, many thanks!
Cheers
Mark
|
|
|
|
|