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

Conference clusta::acms

Title:ACMS comments and questions
Notice:This is not an official software support channel. Kits 5.*
Moderator:CLUSTA::HALLAN
Created:Mon Feb 17 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:4179
Total number of notes:15091

4077.0. "sigh-in transaction delay" by EWBV05::OHSHIMA (Yuichi Ohshima, System Support Tokyo) Tue Jan 14 1997 04:18

T.RTitleUserPersonal
Name
DateLines
4077.1"Flooding the ACC?"CSC32::D_CAREWTue Jan 28 1997 18:0726
        ||>the submitter node is making polling request to back-end 
	||>node using acms$get_procedure_info to see whether the application
	||>is available or not. At this time the polling request is executed 
	||>about 45 times per minute.

	||>While the back-end node was going down, if new user try to sign-in to 
	||>back-end node and submit transaction to back-end node, the transaction
 	||>delay was occurred. 
    
    
    HERE'S MY GUESS:
    $get_procedure_info is not designed to be a polling service executing
     45 times per min-- this action is basically flooding the BE ACC to 
    the point that other ACC business (such as login validation) is being 
    starved out.  This would be especially likely if you have more than
    one FE, or more than one FE agent process or both.  Both $get_proc_info
    and $sign_in services use the ACC -- and the BE ACC is involved in
    distributed configurations.  
    
    Couldn't you use ACMS's builtin primitive failover, (search list
    logical names), rather than such strenuous polling, or at least slow
    down the polling frequency?
    
    Dave Carew
    
    
4077.2OHMARY::HALLBill Hall - ACMS Engineering - ZKO2-2Tue Jan 28 1997 19:4117
    
    	To add to Dave's explanation (thanks Dave)...ACC, like most
    	ACMS processes is AST driven.  Messages are received in
    	EXEC mode and then delivered to the process as a USER-mode
    	AST.  While a process is executing at AST level, it cannot
    	be interrupted at the same level until the AST is dismissed.
    	Since the bulk of ACC works occurs at USER-mode AST level,
    	they just back up waiting for the current AST routine
    	to complete.  There is no threading here, no time-slicing,
    	nothing.  A routine executes to completion, dismisses the
    	AST and the next one gets delivered.  If you bombard the
    	ACC with requests, then you just have to wait for them to
    	complete.  Since most ACMS processes run at the same priority,
    	you could change the ACC priority to something a little higher,
    	so that it could be selected to run more frequently.
    
    	Bill
4077.3thank youEWBV05::OHSHIMAYuichi Ohshima, System Support TokyoMon Feb 03 1997 04:3711
	Thank you for your reply,
	Dave and Bill.

	This customer;Japan Telcom does not use ACMS's primitive failover.
	I don't know the reason of not using this, probably poor design ?	
	
	Thank you,
	Yuichi Ohshima,
	System Support Tokyo.