[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

875.0. "Monitoring Applications?" by HAM::SCHUKAI () Fri Apr 05 1991 11:23

Hallo,

we must present the next steps of integrating DECmcc in a customers Management
Environment. One of the functions they require is analyse the applications. 
That means how can I use DECmcc to obtain data or messages if an the use of 
an application reaches a threshold for line traffic, disk space, response
time etc.
As I know the current version of DECmcc doesn`t support these functions.
But possibly there is an Asset or anything like this available which we can
use to satisfy the customer.
If there is nothing, when will DECmcc be able to support those functions?

Thanks in advance

Dieter

T.R	Title	User	Personal Name	Date	Lines
875.1	If I understand your question...	BSYBEE::EGOLF	John C. Egolf LKG2-2/T02 x226-7874	`Mon Apr 08 1991 22:38`	15
	Dieter, If I understand your question/statement, you want to alarm on thresholds. The current version of DECmcc DOES allow you to do this. You can alarm on a statistical value. For example, if a lines utilization exceeds 80%. We generally refer to applications as functional modules. You can alarm on any value you can do a show on. If you write an AM to retrieve disk space and you have a functional module that calculates percent of utilization, the alarms FM can "send you a message" or start a backup or purge down to n levels when the percent you want to know about is reached. JCE
875.2	The problem is not so trivial	HAM::SCHUKAI		`Tue Apr 09 1991 07:23`	41
	>>>> We generally refer to applications as functional modules. You can alarm on any value you can do a show on. If you write an AM to retrieve disk space and you have a functional module that calculates percent of utilization, the alarms FM can "send you a message" or start a backup or purge down to n levels when the percent you want to know about is reached. Hi John, thank you for answering soon. I agree with the threshold functions and statistics but the problem is much more difficult. What I mean by "application" is not the FM. For example if you run different applications like logistic and pps on the same system. The customer want to know which one causes the most traffic, use n percentage of CPU, use n disk space, have n seconds response time, etc. The next step could be to analyze this on a user/application basis. It is easy if I run dedicated systems - one per application. But it seems to me very difficult if there are multiple applications on one system. Looking forward to distributed computing this question is definitely not trivial. The main problem seems to me to analyze the contents of the traffic, looking which process causes this, assign this to an application and then calculate the statistics. The customer want to use this data for capacity planning, accounting, program tuning, etc. I hope this is a better description of what the customer really wants. As far as I know there is no module or asset available which solve this problem. But have we any idea to solve this and who is working on it? Regards Dieter
875.3	It's not just for Networks anymore...	MCDOUG::MCPHERSON	i'm only 5 foot one...	`Tue Apr 09 1991 09:44`	28
	Hi Dieter. "...analyzing the contents of the traffic" (ala a "Sniffer-like" capability) is not something that we can do right now within DECmcc. That is a future that is being looked into, so hopefully the right person will see your request here and pursue it with you (offline, that is; this is a w:r notesfile). About the "application monitoring": Sounds to me like you're looking for something like VPA (DECcp or whatever we're calling it) to share data with DECmcc... I think that VPA has a set of callable routines, so it may be possible to write some sort of MM that hangs around and uses these routines to scoop up VPA data. You could then fabricate a mechanism to link this information to the Alarms FM / Notification Services piece of DECmcc... Wait. Maybe you don't even need the VPA libraries. I bet you could just write a couple of routines that would scan the system (using regular old System Services) and use that raw system data to help DECmcc generate alarms. There's obviously a little more work to it than that, but that kinda seems reasonable. I think there's been a few VAXsim people lurking in this conference. Maybe they're already up to something like this or they can shed some light... /doug
875.4	Talk to VPA folks	LINSEY::OBRIEN		`Mon Apr 22 1991 18:59`	19
	I agree with -.1, but with a further warning. You are confusing the mechanism with the problem analysis. You shouldn't be worrying so much about event logging per se, but rather threshold monitoring, which can be implemented by RPC callbacks (and which is why they were invented) as well as by posting an event. If you want API code now, you should talk to the VPA or the ISB DECAlert folks, but don't count on the mechanism always being 'post an event', particularly in heterogeneous systems whose components are inevitably each designed by someone trying to solve slightly different problems. <rhetorical flag set> When are we going to learn that the best engineers fit the solution to the problem, not the other way 'round? Ans: When it becomes clear that solutions are invariably apples or oranges or bananas presented as generic fruit. <clear rhetorical flag> Linsey O'Brien Event Services architect