| Author |
Message
|
| eva555 |
 Posted: Tue Nov 20, 2012 1:41 am Post subject: disc read/write for a message via MQ |
 |
|
Novice
Joined: 10 Sep 2012
Posts: 15
|
| I was recently asked on how many times a disc read/write operation occurs from an MQ message from the time an mqput and mqget occurs. I have never read anything on this before and wondering if it was a trick question. Would you pls let me know anything on this???? |
|
| Back to top |
|
 |
| exerk |
| Posted: Tue Nov 20, 2012 1:59 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
The answer may be none for a non-persistent message to at least twice for a persistent message. Who asked the question, and in what context?
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| eva555 |
| Posted: Tue Nov 20, 2012 2:24 am Post subject: |
|
|
Novice
Joined: 10 Sep 2012
Posts: 15
|
It was an interview question  you guys may not encourage interview questions in this forum but this is just out of curiousity.
I did reply that it would have to be twice but then he was not convinced |
|
| Back to top |
|
|
| exerk |
| Posted: Tue Nov 20, 2012 2:28 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
Interviewers that ask 'trick' questions - bad practice on their part. That said, if you think through the logical sequence of events when a message of either persistence is put and got you'll have the answer.
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| markt |
| Posted: Tue Nov 20, 2012 2:31 am Post subject: |
|
|
Knight
Joined: 14 May 2002
Posts: 508
|
| The only correct answer is "it depends". Not just persistence, but size of messages, previous work, any concurrent work, other application accesses to the same queue, transactionality (and probably more considerations too) ... |
|
| Back to top |
|
|
| eva555 |
| Posted: Tue Nov 20, 2012 2:52 am Post subject: |
|
|
Novice
Joined: 10 Sep 2012
Posts: 15
|
| Does it mean that the parameter LogWriteIntegrity in qm.ini does not affect this ? |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Tue Nov 20, 2012 5:51 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
The correct answer is 0, 1 or more times - depending.
All messages are first put to a buffer in virtual storage. If the message is persistent, then it will be written to disk (log). So, one write to disk.
From time-to-time, and at qmgr shutdown, buffer management software will write messages from buffer to disk. So, at least one (more) write to disk.
If the consuming app is alive and well, it will consume the message from the buffer. In this case, for non-persistent message, there will be zero writes to disk.
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| markt |
| Posted: Tue Nov 20, 2012 8:26 am Post subject: |
|
|
Knight
Joined: 14 May 2002
Posts: 508
|
| Quote: |
| If the message is persistent, then it will be written to disk |
Not necessarily. IT DEPENDS. |
|
| Back to top |
|
|
| exerk |
| Posted: Tue Nov 20, 2012 8:29 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| markt wrote: |
| Quote: |
| If the message is persistent, then it will be written to disk |
Not necessarily. IT DEPENDS. |
Would you care to expand on that please, because I understand the way WMQ works to be that persistent messages are committed to the logs (distributed platforms) before being made available to an application? In my book, that's one definite IO access.
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| Mr Butcher |
| Posted: Tue Nov 20, 2012 8:54 am Post subject: |
|
|
Padawan
Joined: 23 May 2005
Posts: 1716
|
| bruce2359 wrote: |
| If the consuming app is alive and well, it will consume the message from the buffer. In this case, for non-persistent message, there will be zero writes to disk. |
How about log i/o for UOW logging if the get is done with syncpoint ?!?
@all
IMHO "It depends" is the very correct answer. You can even think of i/o that is done to the mqerror log files because of a channel start that is caused by a xmit-q triggered channel that is fired because a message is put or other strange stuff like update queue depth counter and so on. theres a lot of i/o, when a message is put or get. not just 1 to buffer and 1 to log....
_________________
Regards, Butcher |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Tue Nov 20, 2012 9:02 am Post subject: Re: disc read/write for a message via MQ |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| eva555 wrote: |
| I was recently asked on how many times a disc read/write operation occurs from an MQ message from the time an mqput and mqget occurs. I have never read anything on this before and wondering if it was a trick question. Would you pls let me know anything on this???? |
It's definitely not a trick question. Your answer will demonstrate your understanding (or lack of understanding) of WMQ fundamentals. In this case, logging.
So, at mqput, a persistent message will be written to the log disk before being delivered to the queue. At mqget, the message will be written to the log disk before being delivered to the consuming program. In the gap between put and get, buffer management software may/will write the message to the queue on disk.
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Tue Nov 20, 2012 9:25 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
Yes,  It depends.
The interview question is somewhat open-ended. As such, it is a good question.
A good question solicits specifics as to how the product behaves, but also allows the interviewer to observe how the interviewee thinks through a problem.
My answer focused solely on the message, and omitted syncpoint information which is also written to the log.
Of course, we can get lost in the definition of an I/O, how many I/Os, etc..
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Tue Nov 20, 2012 10:30 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
An MQPUT with no syncpoint of a persistent message to a queue where there is an outstanding Get With Wait with no syncpoint will have the persistent message pass between apps with no writes to a log or queue file, if I recall correctly.
Whether its a good design to MQGET messages without syncpoint if they are important enough to be persistent is another question... (ah, no.)
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| lancelotlinc |
| Posted: Tue Nov 20, 2012 11:32 am Post subject: |
|
|
Jedi Knight
Joined: 22 Mar 2010
Posts: 4941
Location: Bloomington, IL USA
|
| exerk wrote: |
| Interviewers that ask 'trick' questions. |
I find that companies who allow interviewers to ask meaningless questions are companies I would not want to work for anyway.
Knowing the answer to this question does not help implement a system more efficiently, or in any way impact cost of transactions.
Back in the olden days when disk space and I/O bandwidth were constrained, such information may have helped implement a more efficient system. But with today's relatively unlimited I/O bandwidth and the cost-per-byte for DASD being very low, it's like asking you how much does your long distance cost you on your mobile plan. The answer is pointless since the wireless carriers throw long distance in for free.
http://en.wikipedia.org/wiki/John_Draper
Back in 1970, long distance cost US$3 per minute, it would be worthwhile knowing how many times an MQ message hit the disk.
Phreaking
As Colonel Jessup once said: "I'm a fair guy, but this Phreaking heat is making me crazy." "She outranks you Danny boy. "
_________________
http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
| Back to top |
|
|
| exerk |
| Posted: Tue Nov 20, 2012 11:37 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| PeterPotkay wrote: |
An MQPUT with no syncpoint of a persistent message to a queue where there is an outstanding Get With Wait with no syncpoint will have the persistent message pass between apps with no writes to a log or queue file, if I recall correctly.
Whether its a good design to MQGET messages without syncpoint if they are important enough to be persistent is another question... (ah, no.) |
And there was me thinking that all persistent messages were logged, irrespective of a waiting getter, or syncpoint, or not, because it's not going to be persistent if the box dies. Every day still a school day...
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
|
|
