[Beowulf] So we will write our own book - next steps...

Robert G. Brown rgb at phy.duke.edu
Mon Feb 28 17:28:12 PST 2005

On Mon, 28 Feb 2005, Ryan Sweet wrote:

> > Based on Glenn's comments, I was actually feeling (once again) like I
> > ought to try to shake free enough time to do another full pass through
> > the content to bring it up to date and see if I can finish off some of
> > the missing chapters and -- possibly -- seek a paper publisher.  I want
> > to keep it online/free either way (and there are publishers out there
> > that are comfortable with this) but a lot of people want to own a paper
> > copy of stuff like this.  I get a lot of requests for a printable PDF
> > from people all over who found the html with google but missed the
> > online pdf images right next door...
> Which is another reason that maybe a wiki would not entirely serve (see 
> above).
> OK, this may be a loaded question - for Robert, how do you feel about people 
> contributing to your book vs starting a new, collaborative effort which draws 
> upon the strenghts of what you have already done?

I'm perfectly happy either way, as long as no further demands are put on
my time (which is under a tension of a dozen kilonewtons or so:-).

I put an OPL on the document for a reason -- as long as you don't
publish an effort that contains a whole lot of my writing for money and
not give me any (or get permission to do so ahead of time), you're
welcome to steal, reuse, borrow, adapt, or otherwise mutilate my efforts
in anything you put together with some measure of attribution and the
viral copyleft thing in force.

As I also said, I welcome contributions -- if anybody wants to
contribute chapters that would be great, and I'll even leave your name
at the top of your chapters.  One concept that I had for the book some
time ago that isn't really implemented is to make it a kind of revolving
"journal of cluster computing".  I've written what might be viewed as a
core/intro to cluster computing, with fairly detailed sections on at
least some of the important stuff.

What it NEEDS is somebody who is a Myrinet expert to write a chapter or
article on "Using Myrinet in a Compute Cluster" -- stuff on getting it,
installing it, plugging in the hardware drivers and so forth so that
e.g. MPI can run on top of it, some example programs (toy code or real
applications) that run on 100 BT TCP/IP and Myrinet side by side for
timing and parallel speedup comparisons.

Ditto for SCI.  Ditto for Fiber Channel, infiniband, gigabit ethernet.
An article on diskless clustering.  An article on installing and using
warewulf on top of e.g. RHEL, FC2 or FC3, Centos, Caosity.  An article
on SGE.  All by people who actually use all of the tools in their daily

This is where I, or any possible author, come up short.  I know
something about all of the above, but I don't have direct experience
with all of it and don't know a lot of people that do.  Greg, probably,
and Don.  People in the business side of building clusters so that they
end up hands on with lots of hardware configurations.  A few people in
the REALLY big cluster compute centers.

So what we NEED is some of the real experts on the list to write expert
level but user-friendly contributions.  If these were done as "articles"
rather than chapters per se, it would also address the problem any such
documentation has with information getting "stale" quickly.  Without
updates every year, a lot of the technical stuff has such a short
half-life that any cluster book quickly becomes nearly useless beyond
the intro level.  I haven't done a major catchup on my book for a couple
of whole years, and it is already woefully behind.

Alas, my experience with co-authors so far hasn't been too positive. I
think no fewer than five or six people have offered to do everything
from write half the book with me as a full co-author to contribute a
chapter here or there, and I have yet to see a single line of actual
contributed text.  Hence my cynicism -- we are busy, we are all busy.
Writing is a LOT of work (I promise -- it is one of the things I "do").
Most folks don't realize how hard until they have to write a twenty or
thirty page chapter (maybe with references and figures) that needs to
pass some sort of review and that other people will read and everything.
Twenty or thirty hours later...

Well, apparently they quit before they get to the 20-30 hour mark.

So I will watch with great interest as you try to get something
together, and will applaud your energy and determination if you succeed.
A wiki/blog sort of thing actually isn't such a terrible idea if you can
push it to the critical point where enough people participate and
contribute.  Sort of an online freeform journal.  But then, this list is
(if and as google succeeds in getting to the online archives) already a
pretty hellacious resource in that regard.

> For others, how would you feel about contributing to Robert's book using his 
> Latex template, vs starting a new collaborative effort?
> Each approach has advantages though if, as was mentioned, its been difficult 
> to get wider contributions for the book as it is, then maybe a more overtly 
> collaborative approach would help.


Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu

More information about the Beowulf mailing list