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

Conference ranger::pwosf

Title:PATHWORKS for OSF/1
Notice:see also NOTED::PWDOSWINV5 (PW client) & TURRIS::DIGITAL_UNIX
Moderator:CPEEDY::LONG
Created:Thu Apr 22 1993
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:1874
Total number of notes:6870

1747.0. "Blocked minimum ? realy this long." by KERNEL::BURNST () Thu Feb 13 1997 09:51

    Hi,
    
    Could anybody explain the use of the terms 
    
    SAM db RAMsem statistics:
            has been exclusively locked 1197403 times
            blocked 342 times (blocked minimum 17100 mSec)
                               ~~~~~~~~~~~~~~~~~~~~~~~~~~~
    This is from the lmastat -a on a PW DECUNIX v6.1 eco 1 server.
    Does this realy mean that the MINIMUM time blocked was 17.1 secs
    and therefore a maximum time of greater than this ?.
    The cust is also seeing similar figures to this on the acl db
    statistics.
    Is this the minimum time a user who is blocked will wait ?
    Is this the maximum time a user who was bloccked waited ?
    Is this the total wait time all users have waited ?
    or what !!!!
    
    Trev UK CSC
    
    
T.RTitleUserPersonal
Name
DateLines
1747.1CPEEDY::HORGANFri Feb 21 1997 14:1331
	The SAM data base is a set of files which contain all sorts
	of server administration information.  This data base is mapped
	by all server processes.

	The lock statistics represent an activity indicator, not an inactivity
	indicator.  The time mentioned must be looked at two ways:

		NOTE:  The time value is the product of the number of
		       times blocked times the 'lock nap time', which
		       is usually called the 'back-off' time period.

		1) This is the cumulative time by all waiting tasks which
		   waited on this lock.  

		2) Since the lock is 'locked' only when in use by
		   another process, so this is *** also *** an
		   indicator of amount of time this resource is being
		   used.

	Also note that other tasks in the same process may run when the
	task that 'blocked' isn't running.  So this is *not* an indicator
	that the whole server process is idled.


	Kevin Whitney