[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
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 |
6160.0. "Domain lookups very slow" by DV780::SMITHG () Fri Nov 04 1994 18:46
My customer is running MCC 1.3, VMS 5.5-2 and DNS 2.0. The MCC
domain structure and DNS directories are defined as below. What's going on
is that when they click on certain domains within MCC it takes forever to
complete the "look into" - basically to open up the domain and display it
on the screen. The worst example of this is when opening domain
United_States; we've clocked 18 minutes on the 4500 and 12 minutes on an
alpha we're using for test. I am trying to help the customer figure out if
the problem in with DNS or MCC.
(most of the network entities are DECnis routers and are accessed
via the DECnet/OSI AM; 50 of so of them.)
I'm not at all familiar with how DNS does it's lookups and wondered about
the possibility that the dictionary structure is defined as well as it
could be. I'm particularly concerned about the huge flat structure that
they use for XYZ.MCC. Does DNS index thru something like that or is it
a sequential path that's taken?
BTW: the DNS server is not on the same machine that's running MCC.
I've also looked thru the older notes that addressed DNS performance
and they appear to have been resolved in earlier releases of MCC or
DNS.
Any help you can give me on this would be great.
Thanks, Gary
MCC Domain Structure:
XYZnet
|
--------------------------+--------------------------------
| | | | ...
Canada United_States XYZwan*
|
-------------+--------------
| | | | ...
AZ CA CO *Note: The only entities
| defined in domain XYZwan
---------+--------- are the domains for the
| | |... network cabs (CSPPOP_1, etc)
DENV CSP
|
---------+--------
| | |...
CSPPOP_1 CSPPOP_2 <- Lowest level domain defined
| These are the network cabs
--------+---------- in each city
| | |...
CSP001 CSP002 CSP300 <- Network devices in the cab
such as routers, servers,
concentrators
DNS Directory Strucure:
XYZ:.
|
-------------------------------------------
| |
MCC. USA.
| |
------+-------------------------------- |
| | | | |... |
United_States CO XYZnet CSPPOP_1 DENV |
(No further levels defined here; everything |
except the individual network devices are |
defined here... one long horizontal structure) |
|
|
--------------------------+
|
-------------------------------
| | | |...
DEN CSP SCR ZNR <- city codes
|
--------+-----------
| | |...
CSP001 CSP002 CSP300 <- this is where all the
individual network devices
are defined, such as
routers, concentrators,
servers, etc.
T.R | Title | User | Personal Name | Date | Lines |
---|
6160.1 | additional information | DV780::SMITHG | | Tue Nov 15 1994 16:53 | 25 |
| <<< NOTED::DISK$NOTES8:[NOTES$LIBRARY_8OF4]DNS.NOTE;3 >>>
-< DECdns - Digital's Distributed Name Service >-
================================================================================
Note 1188.2 18+ minutes to look into MCC domain 2 of 2
DV780::SMITHG 18 lines 15-NOV-1994 16:51
-< additional information >-
--------------------------------------------------------------------------------
I have a little more information to add on the configuration and the
testing that we have done:
- first, MCC was running in medium confidence not low. We have defined
the MCC_DNS_CONF logical to low and it made a big difference! The area
that still needs improvement is when looking into an MCC domain that
contains mainly DECnet/OSI nodes (DECnis routers in our case.)
- the DNS directories are defined in a single clearinghouse. That will
change shortly when the customer puts a replica of the root directory
on another machine.
- We have not yet been able to do the clerk traces.
What would help me the most right now, are comments on the customer's
DNS directory structure. Does it look reasonable?
Thanks again. Gary
|