Subject:
Re: Beowulf howto
From:
Kurt ####@####.####
Date:
15 May 2004 02:14:55 -0000
Message-Id: <3.0.6.32.20040514211431.009bd430@pop3.spacestar.net>
Thank you everybody for your feedback. I have addressed most, if not all,
of your concerns. That is the purpose of this open forum: come to an
agreement on a final document, suitable for posting on tldp.org.
----------------
Q: Many of you said I should not have sent such a large email, because some
people are on dialup, and do not want to load up large emails.
----------------
A: I am one of them. I connect to the net with a modem, and after joining
this mailing list, I have to connect to my ISP, and walk away, because I
get between ten and twenty emails a day from it.
But the concern was fair enough; so I will post all my further document
updates to the web site http://lam32767.tripod.com for your perusal.
Note that the web site now has a link to my original document proposal,
which also contained the entire text of the document. Funny; nobody
commented that the email was too large when I submitted it then.
----------------
Q: Some of you said I should add more background on why to make a Beowulf
cluster, and what you can do with one.
----------------
A: The web site contains links to the TLDP Weekly News, which discusses my
new document proposal, and adequately answers this concern. Additionally,
I would say, the intended audience is made of people who already know what
a Beowulf cluster is, what they want to do with it, and began researching
it on the web, which most likely led them to beowulf.org, which contains
scores of links answering this concern. After doing all that research,
they will reach the conclusion that I reached when I decided to write the
document: there is no single document on the web that walks them through
the creation of a Beowulf cluster, start to finish. The purpose of my
document.
----------------
Q: Scientific papers have an abstract that describes the document; HOWTOs
should be no different.
----------------
A: Having a background in computer science and chemistry, I couldn't agree
more. But the howto documents I see on tldp, sunsite, ibiblio, et cetera
do not. I would rather follow the format that I see in the majority of
papers. Refer to my first comment about "open forum". Describe an
introduction that you think would be more suitable. I would appreciate any
input. However, I believe that my brief introduction, though not entitled
"abstract", does outline what the document contains.
Additionally, the howto referenced in my document does not contain an
abstract.
Hey Tille [Machtelt], can you offer input? What do you currently tend to
expect for introductory abstracts?
----------------
Q: Links were lost in the text-only document
ssh wolfnn was more difficult to read than ssh wolfNN
----------------
A: Because you now refer to the website, links are valid, and italic
characters are more visible.
----------------
Q: Can you supply a checklist on how to fix the problems that may arise?
----------------
A: I considered it but concluded that the document, while already rather
large, could get way too big because of all the things that can go wrong.
Going through the process of building my own cluster from just the
available documentation on the web took me qutie a while, because of each
new obstacle to overcome. I could not have truly documented EVERYTHING.
Each single subject has a fairly adequate set of documents out there that
are helpful; I am bringing all the pieces together.
----------------
Q: Where do you get lam? What is mpi? Why do you use NFS?
----------------
A: Refer to website; these good questions are now addressed.
----------------
Q: There is another linux distribution that does [..........]
----------------
A: Yup. You forgot to mention Mosix. After I get this howto completed,
I will be moving on to that.
----------------
Q: What about an example program to test the stuff?
----------------
A: I did not examine the other message passing interface software other
than LAM, but I was lucky enough that LAM DOES include an example hello
world program. Sterling's book also contains source code for hello world.
And as you see in the document I do use that to test my install. I really
hope the other message passing interfaces include a simple hello world
example program for testing because it was instrumental in proving my setup
was correct.
Again, everybody, thank you for the input. Let's look at it again, and
further shape the document into something agreeable to all.
Just to make it clear - the website http://lam32767.tripod.com contains a
link: "My current document" that is my latest updated document that
addresses your concerns.
