James,

If your shadow system is not currently limited by CPU or DISC I/O's then 
you might want to look at using our version of NBPOST that supports 
multiple post processes. If you are interested in getting this version or 
in getting help in setting this up please contact us at 949-754-8000 or by 
sending an email to [log in to unmask] 

Here is a short summary of the changes that we made to NBPOST:

"An optional feature of the NetBase Posting process was developed to 
address certain performance problems that some customers encountered when 
Posting was consistently backlogged on a shadow machine.  Provided that a 
shadow machine was not CPU or IO bound, Multi-threaded Post was developed 
to improve posting throughput.  This option, if enabled, increases net post 
overhead, of both CPU and IO, so should be used only when benefits are 
clear.  For instance, when a small subset of Post IO is for a particular 
file or database, that file/db can be assigned a post queue so that it 
doesn’t have to wait in line for all the other posting to take place.  
Another possibility, is to assign post queues by volume, so all shadowed 
files and databases for a particular volume are assigned the same queue.  
If the shadow machine is IO and CPU bound, multiple post queues should not 
be enabled.  Multi-threaded post will run more slowly than single post on a 
machine that is not backlogged since more processing takes place.
 
Basically, multi-threaded post causes posting for a specific node to launch 
a “father” post process that creates up to 100 “child” post processes that 
can run simultaneously.  The POST=p parameter, where p is 0 to 99,  is used 
in the SHADOW statement in NBDIR to assign a post queue number to a file or 
database.  Post child processes are created for each of these post queue 
numbers.  A post queue number 0 is the default if not specified on a shadow 
statement.  

The father post process will act as a dispatcher and send transactions to 
each child post process based on the post queue number defined in the 
shadow statement.  A new set of queue files of the format PCnnnppp is 
created by the father and read by the appropriate child where nnn is the 
shadow node number and ppp is the post queue number from the shadow 
statement.  The PCnnnppp queue files are first built when a record is 
posted to a shadow file, and the nbpost program sees the “POST=p” in the 
shadow entry in the NetBase directory.  Any time post is stopped and 
started again, Post will automatically start the post child processes for 
all PCnnnppp files present."

Dan Clifford
Quest Software 

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