Re: UP Made Distro

Top Page
Author: Dean Michael Berris
To: Paolo Alexis Falcone, ilug
Subject: Re: UP Made Distro
Paolo Alexis Falcone wrote:
> Here's the catch with the thesis - it's theoretically UP's property
> (binary, source code, documentation). although you can exert copyright
> over it, you can't privatize it to be closed-source (which is a good
> thing :-) as it's theoretically not your own intellectual property (the
> exact state of your thesis upon submission). You can create derivative
> works though, pero it would be subject to the IP guidelines of UP - and
> UP's known to implement it fiercely.

you guys know me, (and to those that don't...) i do not have plans of
privatizing linux related stuff i do. although of course, when time comes that i
have to (i.e., work for Microsoft, Sun, or IBM... (wish ko lang!)) then i'll
cross the bridge when i get there. :)

and besides, i love UP. i'm not worried that the work I do will be UP's
property. after all, it's my thesis... ;) having my name there is quite enough
for me. :)

>>I'll be calling it Oblation Linux, and the first release will be dubbed
> "Activist"
> On what architecture would that run? Sparc32 or Sparc64? If so, please
> tell the specs - interested akong i-test yan. If that would be GPL'd,
> even better.

initially, my testing environment is a collection of UltraSPARC 5's -- which i
think are 64 bit machines. however, the distribution i'll be coming up with, the
live CD Oblation Linux distro, will be released to the UP demographic configured
for x86 archs. i will be making a live CD for the Sparc4u's (?) -- i'll be
announcing it when i'm done with it so that i could give interested people a
copy. :)

oh and yes, it will be GPLed. i've consulted my adviser regarding this, and when
i brought the idea up sabi nya:

"if you don't want to have problems regarding UP and licensing, put it in the
public domain before you even submit the thing..."

quite clever i may say is Dr. Albacea... :)

> Try investigating these applications or topics. Some of them are
> existing software that are used already to maintain high-availability
> clusters.

i've seen through literature regarding NUMA, PVM, and OpenMosix. but basically,
i ended up with MPI because i want to limit and simplify the implementation of
the algoritm i proposed.

with NUMA, it'll work on certain hardware (NUMA hardware made by IBM, HP, et al)
and not necessarily on clusters. however, if you have a cluster of NUMA
machines, then you should be darn happy... :)

Heartbeat, AFAIK and IIRC, is a software solution for HA clusters -- those
systems that need to stay virtually up all the time. This methinks is not for
parallel computing...

OpenMosix is a kernel based approach to parallel computing, where nodes part of
a OpenMosix cluster work as one in the kernel level. it basically makes the
linux kernel a distributed linux kernel which has the ability to do message
passing and rpc in the kernel level. this offers automatic load balancing, and
the familiar and intuitive fork-and-forget approach to parallel computing
systems. This is a very elegant solution to the problem, but if i had to modify
the linux kernel and implement my load balancing, i might as well stop studying
and go work some place NOW. (if you know what i mean...) ;)

> * Piranha

first time i've heard of this... me should luk. :)

> There are other components you should look in to though... just research
> on it further (as literature on it is fortunately abundant).

unfortunately i have stumbled upon a wealth of information regarding clustering
technologies -- there's condor, and some other freaky ways of doing parallel
computing. however, i am more interested in studies which actually did some load
balancing work, or different (read: non-conventional and experimental) means of
processing data in a distributed parallel machine.

the literature for the latter, is unsurprisingly sparse. i assume that people
like their lives easy and livable instead of complicated and boring. ;)

however, there was literature available and studies (not surprisingly) from
universities and colleges all over germany, the US, Japan, and some even from
Spain are those that i have cited. however, the most striking paper would be by
Dr. Luis Sarmenta of ADMU about bayanihan computing -- his Ph.D.
dissertation/thesis (?) at MIT : that's Massechussets (? did i spell that
correctly ?) Institute of Technology.

i wonder why UP is lagging behind in research...

>>it is due to come out some time this July, hopefully before my birthday. :)
> Hopefully :-)

yes sir, hopefully. :)


aka Dean Michael C. Berris
home +63 49 5680142
work +63 49 5680024
pgp key ID = 0xF9501761

Version: 3.12
GCS d- s:+ a--- C++ UL++ P+ L+++ E- W+ N* o K w--
O---- M- V-- PS+ PE Y+ PGP++ t+ 5 X+ R+ tv+ b++ DI+ D+
G e h! r+ y+

Isko't Iska GNU/Linux Users' Group Mailing List
ilug@??? (#ILUG @
Searchable Archives: