Re: High Volume Results Management

I agree.
I was not going to respond because I did not know if there was a new and better 
way than how I do it.
But what I do is to arrange a multi-level hierarchy, so that no enumeration ever 
has to be large.
For example, if you need to monitor the status of 50,000 drives, you establish 
255 managers that each monitor 255 drives and return the results in a single 
chunk of data.
Or you can simply create a lot of subclasses, and only enumerate one subclass at 
a time.
Enumerations seem to slow geometrically with quantity, and this appears true 
with all CIM implementations I have used, not just Pegasus.

 Kirk Augustin
11821 NW McNamee Rd
Portland, OR 97231




________________________________
From: Robert Kieninger <[email protected]>
To: Imran Ansari <[email protected]>
Cc: [email protected]
Sent: Wed, October 13, 2010 11:45:05 PM
Subject: Re: High Volume Results Management

Hello Imran, 

returning a high volume of instances has been a long standing problem with CIM 
enumerate operations  in general.  

To address this problem the CIM operations have recently been extended with the 
pulled enumeration operations, but these have not yet been implemented in 
OpenPegasus nor has the CMPI interface been extended to accommodate them. 

One way to address this is problem is making use of http chunking, what allows a 
client to start processing the returned instances chunk by chunk while the CIM 
server is still sending data. You can use http chunking for example in 
combination with the CIM Client for Java from the SBLIM project on SourceForge. 


Other ways to retrieve partial data would be through some kind of associations 
or using WQL queries through execQuery, if your version of OpenPegasus supports 
this.

Mit freundlichen Grüßen / Kind regards
Robert Kieninger 



From:  Imran Ansari <[email protected]>  
To:  [email protected]  
Date:  13.10.2010 20:12  
Subject:  High Volume Results Management 
________________________________
 


Hello Everyone, 
I've a CMPI based provider. Presently, I'm having problem with the response time 
while returning large number of instances. I would like to know if and how it is 
possible to deliver these instances in an incremental manner via callback or 
some other mechanism? 

  
I would really appreciate some help on this. 
  
Thanks,
Imran