MEDIAW - Media wait - looking for info

[John22439]

Good morning all,

I've come across this a few times in the past, but did not have time to investigate it.

It appears as though the amount of time my backup of test_001, took less time then the amount of MEDIAW that is suffered.

Can anyone explain how TIVOLI gathers or generates MEDIAW or COMM_WAIT data (I also noticed there was a long COMM_WAIT.... is it related?)

Here is the SELECT I used to extract the data:

select * from summary where MEDIAW>300 and (time(end_time)>'15:00' or time(end_time)<'14:59') and end_time>current timestamp - 1 days

START_TIME: 2007-07-30 06:07:16.000000
END_TIME: 2007-07-30 06:50:01.000000
ACTIVITY: BACKUP
NUMBER: 109690
ENTITY: test_001
COMMMETH: BPX-Tcp/
ADDRESS: xxx.xxx.xxx.xxx:xxxx
SCHEDULE_NAME: LVL2_0600
EXAMINED: 31198
AFFECTED: 222
FAILED: 0
BYTES: 13011575251
IDLE: 122
MEDIAW: 3434
PROCESSES: 2
SUCCESSFUL: YES
VOLUME_NAME:
DRIVE_NAME:
LIBRARY_NAME:
LAST_USE:
COMM_WAIT: 1516
NUM_OFFSITE_VOLS:


As you can see the backup started at 06:07:16.000000 and ended at 06:50:01.000000 for a duration of approx 43 minutes.

The Media Wait was 57 minutes??? How is this possible? (is the COMM_WAIT a contributing factor?)

Any help would be appreciated.

Cheers,
j

 

[BBB]

You have "PROCESSES: 2". The real world time was 43 minutes but 2 processes between them effectively get 86 minutes of total time to do stuff, which is recorded into the mediaw and comm_wait values as a combined total. From rough calculations, its possible your 2 processes spent 28 minutes each of the 43 minutes total time waiting for media.

The comm_wait isn't really related to mediaw at all.

 

[John22439]

Thanks for the quick response BBB,

Ok, but my understanding was that one of the sessions was to intiate/authenticate the connection between the client and the server, and the other session was for the actual data transfer. I guess my next question is are PROCESSES the same as SESSIONS?

If so, when ever I've seen this live, the MEDIAW only showed up on one of the sessions, the other (the "initiator") was in and IDLEW state. If I understand you correctly, you are saying that both sessions affect the MEDIAW - so the "actual" media wait time is only half of whatever TIVOLI reports?

I hope I'm wrong.... ???

j

 

[BBB]

Hmmm I was assuming you have resourceutil set to 3 or thereabouts. I'm not sure if the control sessions show up in the "Processes" entry. Don't have access to a server to check, but from memory my clients have resourceutil set to 2 and "Processes: 1" in the events table but could well be wrong. Would have to check...

Perhaps both the data and control sessions count their time as mediaw while waiting (with resourceutil 2)?

 

[John22439]

Well,

As far as I've read in the RED books for the client version we are running, and since the "resourceutil" is not explicitly set, it defaults to 2. So our clients have their resourceutilization set to 2 automatically.

As for if TIVOLI counts both... I really don't know, and hope this is not the case.

j

(How can this be confirmed?)