[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference azur::mcc

Title:DECmcc user notes file. Does not replace IPMT.
Notice:Use IPMT for problems. Newsletter location in note 6187
Moderator:TAEC::BEROUD
Created:Mon Aug 21 1989
Last Modified:Wed Jun 04 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:6497
Total number of notes:27359

5594.0. "MIR capacity" by FRAIS::WELLERSHAUS (Heidi Wellershaus) Tue Sep 14 1993 11:07

We are working the first time with decmcc (FT1.3) on ULTRIX 4.3 in a project 
for the German Telekom. Storing a large amount of data of object instances,
we made the design decision to use the private MIR (implemented with nDBM)
to take advantage of the MCC_MIR routines. 

People at VBO told me about a MIR capacity limit of 100.000 records of up 
to 4.000 bytes.

Because we don't have any experiences with decmcc, I have the following 
questions:

- who knows more details about this limit?
- where can I find more information or documentation about nDBM restrictions?
- did anybody use the private MIR to store a bulk of data?
  If so, what kind of limits occured (number of objects, number of repository)? 


Working with the private MIR (writing and reading instances and/or attributes)
the following error message occurs aperiodicly on the screen:
mcc_ndbm bad item size = 805308416

I don't use the mcc_ndbm routines, only the mcc_mir routines.

Thanks for any hints, 
			Heidi
T.RTitleUserPersonal
Name		DateLines
5594.1Partial answersTAEC::FLAUWFri Sep 17 1993 05:4837
re: .0

    Heidi,

    
    I didn't answer your note earlier as I was hoping that somebody from
    EMF Eng. might be able to comment on your base note.

    As we (TeMIP) are part of the "people in VBO" you mentionned in your
    note, I would like to add a few comments on your note.

    For TeMIP V1.0, we used the nDBM package for our internal datastore
    (not using the mcc_mir routines). While testing this package, we found
    out that we were getting "split overflow" errors somewhere over 100 000
    records.  We didn't explicitely tested the mcc_mir routines up to this
    limit, as we were not using them for storing events.

    As the mcc_mir routines are using the same nDBM package as the one we
    used in V1.0, they should have the same limit.

    In TeMIP V1.1, we decided to switch to a different datastore package,
    C-ISAM due to nDBM limitations.

    Regarding getting more information about the nDBM tool, I think that
    one of the Ultrix notefiles should the right place.
    
    Sorry of not being able to be of more help regarding your questions,
    but I would suugest that if it is critical for your application to be
    able to store large ammounts of data, you run a small test, writing
    *and* reading back the data in the MIR to check if the MIR can handle
    the volume of data needed for your application.

    Best regards,

    Marc.
5594.2I thought 10,000 was suggested max?BAHTAT::LZOPCB::bondThu Oct 07 1993 11:1610
About 1 year ago, I asked (I think in this notesfile) whether registering 
2,000 entities in the Local MIR was ok because I think you will find that the 
documentation used to suggest 1,000 max.  The response I got was that 2,000 
was fine; performance had been ok at around 10,000 but above this, 
engineering thought that the performance would tail off and you would be 
better off using DNS.

Perhaps things have changed now but that was the answer then.

chris