[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

2469.0. "Nb of Historian background process" by TAV02::KALAI () Tue Mar 03 1992 09:22

Shalom,
I need some advise build upon experience and or knowledge about the best way
to organize the collection of the history.
I have a network build from 20 domain for monitoring/alarm/notification 
with around 700 partition all together.For history purpose I  divided them 
into 5 other domains.
I wonder what is the best choice ,to have 2 big domain just for the history 
purpose or to use the same 20 domains that I use for monitoring/notification.
I think I read somewhere that each background process can handle up
to 500 polls .
any comments?

Regards
Kalai Yael
T.RTitleUserPersonal
Name
DateLines
2469.1an experienceICS::WOODCOCKWed Mar 04 1992 00:5035
>> I have a network build from 20 domain for monitoring/alarm/notification 
>> with around 700 partition all together.For history purpose I  divided them 
>> into 5 other domains.
>> I wonder what is the best choice ,to have 2 big domain just for the history 
>> purpose or to use the same 20 domains that I use for monitoring/notification.
>> I think I read somewhere that each background process can handle up
>> to 500 polls .
>> any comments?

I'd be interested in seeing what the interval was for the 500 polls and on
what size system. In other words is it 500 polls per hour for a non-loaded
4000 (or some other system)

I have just gone thru some 'mcc' tuning very similar to your question. I had
12 domains set up in the monitor for alarms/notification. I then simply used
these 12 domains for my historical polling. It had actually run for several
months without too many complaints. But when performance began deteriorating
I found these 12 jobs growing individually to 20-30% each at times. The 8810
was dragging with 100% cpu utilization. I had 75 partitions polled twice an
hour and 150 partitions polled once a day.

I reconfigured with 4 domains and cut the original 75 partitions to once an
hour. We are now humming at 30-40% cpu. I definitely do not recommend 20
domains as these processes consume lots of resources. I do not know if 2
domains will be enough for you. You state 700 partitions but how many per
hour with what sized system? Counters and status would be the only two 
needing any frequency. The rest can be once a day. I believe I may have room
to grow inside my current four domains but I don't know how much.

Something which may be of equal importance is disk i/o. I was also bottlenecking
in this area at one point. Reducing the poll processes keeps the number of
open files down and we even moved all our log files to a different disk.

best regards,
brad...