RE: [SAGE] Enterprise Documentation

[John LLOYD <jal@mda.ca>]

> From: Craig Hancock [mailto:chancock@delphi.bsd.uchicago.edu]

>  Anyway after upgrading the
> machine I had the pleasure of switching the machine over to 
> gigabit. During the course of the week
> I had a lot of performance problems. The load average on the 
> machine would spike to 9.00. I didn't
> understand why the load was so high it was never high before. 

You didn't say much about the observed performance problems.  If the
throughput was lower, or  the file access failed, or what?  Were big file
transfers working and little ones getting slower? 

The load average is not a symptom, it is an observation....be wary of trying
to "fix" non-problems.  Really, do your users care about the load average?
No, they want fast file access.


> I attempted to check
> 
> 1) Network throughput
> 2) IO throuput 
> 3) Avaliable memory
> 

A good list!

> ... and I realize I am lacking in
> 
> 1) Unix kernel heirarchy
> 2) Device drivers
> 3) File system structure
> 4) Newtwork issues
> 

Another good list!  You are ahead of a lot of people because you have an
understanding of what some of the potential issues are.  Sort of like you
know what it is you don't know.

Nobody necessary learns all about these things first---and then decides they
are ready to solve problems.  Most of us learn from the same situation you
find yourself in---a problem is discovered and you have to dig into things
to fix it.

> Is this stuff you learn from expierence or from a book. I 
> think I have reached a cross roads 
> where howto's and books aren't enough anymore for me to 
> learn. 

Somebody somewhere is always writing a new book...

Some of the book on O'Reilly's list are good: High Performance Computing, by
Dowd and Severance.  Some of the stuff Sun publishes on Solaris (Adrian
Cockcroft), and Oracle (standard docs plus some of the Osborne titles) even
if you don't use Oracle they tell you lots about how to configure things.

A good one, hard to find, is The Art of Computer Systems Performance by
Jain.

Experience with other operating systems helps provide some context (linux
versus solaris is not quite the same thing as comparing linux and windows
95....both comparisons have something to tell you.  Of course when I say
windows 95 vs linux I mean as a personal system.)

> to go to get a deep understanding of such concepts. I guess I 
> should ask how does one begin
> to understand such topics?
> 

The School of Hard Knocks.  No address available.


-John