[SDBUG] January Meeting

[James, Jay]

Google does things differently then most would expect. Even differently
than most average IT people would expect. 

Its not load balancing for the sake of load balancing, at least not in
the traditional sense. Think more 'massive parallelism' and less 'load
balancing'. Sheer scale simply allows availability. 

Instead of a single query being executed on a single system against the
entire database of crawled websites, the query gets sent to and then
gets executed on hundreds of systems known to be available, where the
results become aggregated by a 'team leader' mechanism, finally shipped
and presented to the querying party on the Google results page.

So the concept in reference to load balancing and Google may not
entirely be apropos, where in actuality the available computing
resources are a known quantity before the search is committed, and the
routing is done that way.
Hm, its not load balancing per se. Its grid balancing.


Anyway, this is an extremely rudimentary breakdown and is a pretty fair
assessment as of the summer of last year :) 


Jcj


Maybe I am nit picking with semantics :) if you feel I am, excuse me.
Its Friday and I am itching for the weekend!






-----Original Message-----
From: sdbug-bounces at sdbug.org [mailto:sdbug-bounces at sdbug.org] On Behalf
Of Peter Leftwich
Sent: Thursday, January 04, 2007 10:56 PM
To: SDBUG
Subject: Re: [SDBUG] January Meeting

On Thu, 4 Jan 2007, Michael J McCafferty wrote:
> Yes... we are still here. Finishing up the drinks. Done with food. 
> Waitress Christina looks radiant tonight. I am reading her this email 
> and she is shaking her head, telling me I am crazy.

I was there tonight.  She looked hella radiant, but it didn't take (or 
shouldn'ta taken) my psych BA to quickly deduce (as she owned up to)
that 
it was a "new" "guy" in her life, that was making her appear so.  *grin*

Maybe I am getting wise, instead of getting old.

> Topics of discussion this evening:
> Why not to use RAID 5 for databases (relational or otherwise) or any 
> random-like IO.

I remember when I first saw the output of the DOS command "dir" on a 
server running MS Exchange.  I was on it to do some tape rotation.

"SOMEBIGHUGESINGLEFILE.EDB" 98123752398572389275 KBytes (or something)

or something like that... it was one huge file!  the whole server was
one 
...huge database file?  I couldn't believe it.

So tonight when we were talking about RAID 0 1 4 and 5.00006 :) I still 
find it amazing that for example, MS Exchange can run on 3 HDD's and you

can just suddenly PULL OUT ONE HDD and everything goes on fine
unaffected!

> LUN and FS size limitations and experiences with various OSes and
SANs.

When we talked about load balancing, I meant to ask - with a huge
website 
such as yahoo or google, where is the computational power "happening"
that 
is constantly monitoring something similar to "top" or "ntop" so as to 
*know* where to reroute balancing to?  (I am dumbing this down
undoubtedly.)

> How to transfer several terabytes from one system to another over a
LAN. 
> Encountering issues with 10's of thoudands of files copying from
Windows 
> to RedHat Enterprise server.

Would that fit on the USB thumbdrive that's in my wallet?  ;) Not til
3010.

> Dogs.

Yes we talked about a dog.  You did.  *smile*

Anyone reading this who hasn't come to a monthly, you should, it is 
dinner, and social and unstructured and good-natured.  Peace, propz.

--
Peter Leftwich, Owner
Video2Video Services
Box 13692, La Jolla, CA, 92039, USA
http://Www.Video2Video.Com
_______________________________________________
SDBUG mailing list
SDBUG at sdbug.org
http://lists.sdbug.org/mailman/listinfo/sdbug