Hi David, thanks for the reply.
I have tried to apply the netsnmp_register_table() as you suggested in
previous email.
However the table_handler was not called for any Get Request for a table
entity. It was called for a GetNext with a junk request. I used a debugger
to verify this. So I was puzzled and decided to improve the table iterator.
However I'll be happy to use a much efficient method.
I don't know if this is too much to ask... I'll be thankful if you can
verify that this is working on your side using a simple table. There might
be a detail left out...
I'll be happy to try it again.

Thanks again,
Shuki

-----Original Message-----
From: Dave Shield [mailto:[EMAIL PROTECTED] 
Sent: Friday, October 07, 2005 11:15 AM
To: sasson, shuki
Cc: [email protected]
Subject: RE: Cannot find an adquest mib2c configuration file for travesing
large tables. --- Does Net-SNMP fit to real time environment?

On Fri, 2005-10-07 at 10:48 -0400, sasson, shuki wrote:
> Hi all, I managed to "create" myself a pass through interface to traverse
> big tables. Basically it is based on table iterator  template.

> Even after applying this fix I see 35% degradation in performance
> compares to another agent implementation. 

Without seeing the code that you've developed, it's impossible to
comment as to whether this is typical.   But I would be extremely
suspicious about the efficiency of *anything* based on the iterator
helper.

The fundamental idea of this helper was to hide all the OID instance
processing from the MIB developer.  To this end, simplicity was put
above efficiency.   This is easily the least efficient helper
framework that we provide.

Comparing this (or anything based upon it) to the performance of
another agent (presumably using a more efficient processing model)
is more-or-less meaningless.


> Does Net-SNMP fit to real time environment or is it too heavy?

It's probably fair to say that the Net-SNMP is not the sleekest
agent toolkit on the marker.  If nothing else, there's a whole
lot of cruft in the main SNMP library, which has been developed
piecemeal over the years - much of this may well be less than
optimal.  Plus the agent framework is aimed more at flexibility
rather than sheer speed.

  But I would expect that a more efficient approach to this
particular MIB (even within the overall Net-SNMP framework)
ought to get closer to the performance of other agents.


>  Does anyone have any idea how to improve Net-SNMP performance
> for table handling?

Basically the suggestion I offered last week.

Avoid the overhead of using the specialist table helpers,
and work with the basic 'netsnmp_register_table()' interface.
Most of the helpers are concerned with taking work away
from the MIB developer, and handling standard processing
tasks internally.  But this comes at a price - if you want
to improve efficiency, you need to be prepared to do more
of the work yourself.

Dave


-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
Net-snmp-coders mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders

Reply via email to