HP3000-L Archives

November 2008, Week 3

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
James Hofmeister <[log in to unmask]>
Reply To:
James Hofmeister <[log in to unmask]>
Date:
Wed, 19 Nov 2008 12:57:42 -0800
Content-Type:
text/plain
Parts/Attachments:
text/plain (134 lines)
Hello Paul,

I could stand corrected (I am only a MPE network puke after all)... but one
might say the same of MPE/V vs. MPE/iX.  From my recollection the MPE/V
machines accessed memory in smaller chucks and made much more efficient use
of memory (because there was much less memory to index) than MPE/iX where we
were accessing a much larger amount of memory in larger page sizes.

This may be an argument of efficiency vs. scalability.

Linux of course also has to be everything to everybody (one could say that
MPE focus was online transaction database machine).  There are
concerns/complaints about the Linux kernel and the tug-o-war between the
desktop and server maintainers from a kernel tuning/optimization
perspective.  We have to keep in mind Linux systems scale from desktops with
little 1GB memory to superdome servers with Terabytes of memory yet for the
most part they use the same kernel.  Likewise the Linux file system(s) scale
from MB to EB.  Finally CPU's scale from single socket, single core to
"many" sockets quad core (and more to come).

When a single kernel scales this range of hardware, I would expect
inefficiencies to be seen on the low end.  There are some options though...
32 bit kernel for <8gb desktops, hugemem 32bit kernels for >8gb and 64bit
kernels.  There are also tuning options for IO Schedulers, and we also have
RealTime kernels being introduced now.

Just my .02$

James





-----Original Message-----
From: Paul Raulerson [mailto:[log in to unmask]] 
Sent: Wednesday, November 19, 2008 12:10 PM
To: James Hofmeister
Cc: [log in to unmask]
Subject: Re: [HP3000-L] OT: HP's Newest and Hottest DB Server Offering

 LOL! I bet they are, what with all the people vying for time on the
beasties!

The one big problem with Linux is that once you get way down under the
covers, it just isn't very efficient. 
MPE, VM, ZOS, etc. all blow away Linux in terms of efficiency and the way
they use the hardware, 
and yet the hardware is so darn fast, you don't really notice how
inefficient Linux really is. 

Not that I am complaining mind you, Linux is amazing. But how fast the
hardware is is knock-your-socks-off 
amazing. :)  


-Paul

On Wednesday, November 19, 2008, at 12:01PM, "James Hofmeister"
<[log in to unmask]> wrote:
>Hello Paul,
>RE: [HP3000-L] OT: HP's Newest and Hottest DB Server Offering
>
>These XC-HPC compute clusters are kewl...  some of the ones I have seen are
>greater than 1000 nodes.
>
>I don't get to fix them :)  I usually only see the issues on the backend as
>they tend to be abusive of storage which is when I get the call.
>
>Regards,
>  James Hofmeister.
>
>
>
>
>-----Original Message-----
>From: Paul Raulerson [mailto:[log in to unmask]] 
>Sent: Tuesday, November 18, 2008 11:00 PM
>To: James Hofmeister
>Cc: [log in to unmask]
>Subject: Re: [HP3000-L] OT: HP's Newest and Hottest DB Server Offering
>
>Kind of along those lines, I had the opportunity yesterday to tour the  
>data center
>up at Baylor University, and one of the more interesting things they  
>had up there
>was a supercomputer - from HP! Running Intel nodes and Red hat Linux  
>no less!
>
>http://insidehpc.com/2007/09/27/baylor-university-purchases-hp-cluster/
>
>I was a bit impressed, but it led t me to wonder if MPE/ix and HP3Ks  
>have ever
>developed clustering capabilities and were tey ever used in great big  
>computing
>tasks?
>
>A casual google search has has basically turned up nil results.
>
>-Paul
>
>
>On Nov 18, 2008, at 11:40 PM, James Hofmeister wrote:
>
>> Hello Peter,
>>
>> http://www.oracle.com/features/hp/exadata.html
>> Wouldn't it be nice to have MPE sitting atop this platform with Image
>> intrinsics rewritten to access Oracle DB?
>>
>> Nothing special... This is a rack of standard x86 Linux servers;  
>> lots of
>> internal disk.  The O.S. is Oracle's (OEL) Oracle Enterprise Linux.   
>> If you
>> have had a positive support experience with Oracle with their  
>> applications
>> in the past, then this may work for you.  Support for all software,  
>> kernel,
>> db, etc on the machine is from Oracle.
>>
>> Regards,
>>  James Hofmeister
>>
>> * To join/leave the list, search archives, change list settings, *
>> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
>
>* To join/leave the list, search archives, change list settings, *
>* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
>
>

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

ATOM RSS1 RSS2