| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| OS390 pageset usage |
« View previous topic :: View next topic » |
| Author |
Message
|
| tjfunny1 |
 Posted: Wed Sep 04, 2002 10:22 am Post subject: OS390 pageset usage |
 |
|
Apprentice
Joined: 17 Jun 2002
Posts: 35
Location: Atlanta
|
We are seeing a weird thing concerning pagesets. I did 3 tests to be sure. Local queue, persistent, only queue defined to pageset.
Footnote: I used a browser to verify size of data put on queue.
Test 1:
Put 102 messages (size 23 bytes each) to a queue (pageset exclusive). After end of job, I displayed usage:
CSQI018I !MQS1CPF CSQIDUSE PAGE SET 18 HAS 1078 DATA PAGES. 975 ARE
UNUSED, 103 HOLD PERSISTENT DATA AND 0 HOLD NONPERSISTENT DATA
CSQI030I !MQS1CPF CSQIDUSE PAGE SET 18 USED 0 SECONDARY EXTENTS
BEFORE THE RESTART, AND HAS SINCE BEEN LOGICALLY EXTENDED 0 TIMES
As you can see, it used 103 data pages (guess one is used as a index). According to the MQ docs; each message adds 324 bytes for the MD, 48 bytes (internal), the result is what is used to calculate number of messages per page. In the test above, 23 user bytes equates to 372 bytes per message. Usable space per page is 4057 bytes. 4057/372 = 10.x messages/page. Using this calculation, the test would only use 12 pages, not 103.
Test 2: user data = 100 bytes (102 messages)
100+324+48 = 472 bytes 4057/472 = 9 messages/page
but test shows we used 103 data pages, same as test 1
Test 3: user data = 1000 bytes (102 messages)
1000+324+48 = 1372 bytes 4057/1372 = 2 msg/page
but test used 103 data pages again!
When we defined the pageset, we used the OS390 DEFINE CLUSTER command as follows:
DEFINE CLUSTER -
(NAME(CSQ1.PSID18) -
RECORDS(1000 500) -
LINEAR -
VOLUMES(SYYML6) -
SHAREOPTIONS(2 3) ) -
DATA -
(NAME(CSQ1.PSID18.DATA) )
this results in creating 1078 pages as primary (guess it rounds up to nearest track). This also equates to 6 cylinders/primary and 3 cylinders/secondary extents. or 372 cylinders fully extended.
My question is whether the term RECORDS, relates to a PAGE? I found a item is some IBM doc (believe it was a MQ) that equated RECORDS to PAGES. Also found a note that 1 cylinder of dasd = .733 mb of space (733,000 bytes), or 179 pages.
The pageset can be extended 123 times the secondary extent.
What is the maximum settings for the DEFINE CLUSTER command that would equate to 4 GB?
If this is normal, how can MQ store 999,999,999,999 messages of any size on a OS390 Qmgr? (372 bytes overhead per message) |
|
| Back to top |
|
 |
| blowbeat |
| Posted: Fri Sep 06, 2002 1:35 am Post subject: |
|
|
Apprentice
Joined: 02 Apr 2002
Posts: 49
|
I cannot help much. I think I can only explain why there's 103 pages used when there are 102 messages put. The RBA of the last checkpoint taken is stored in page 0 for recovery purposes.
I'm not a VSAM expert, but I believe it works like this.
Default CI size for a linear data set is 4096. According to the IDCAMS manuals : space is allocated for a linear data set with the number of control intervals equal to the number of records. The default recordsize is 4089. The maximum setting for RECORDS is 16,777,215.
It looks like a new CI is used to store each message in. If you'd like to do some more testing, see what happens if the message size exceeds 4089..
Cheers, Jan. |
|
| Back to top |
|
|
| tjfunny1 |
| Posted: Wed Sep 18, 2002 11:40 am Post subject: FORMAT PAGES |
|
|
Apprentice
Joined: 17 Jun 2002
Posts: 35
Location: Atlanta
|
Sort of, anyway
I created a new pageset, with 100,000 data pages primary, 8000 pages secondary. I also added
to the FORMAT command:
Previous format =
New format =
| Quote: |
| FORMAT PAGES(100000) |
I altered the queues to use the new pageset and now it seems to put more than one message per page. We tested using messages sized 250 bytes and it puts 6 messages per data page.
I used the formula from the MQ docs to verify.
| Quote: |
| 324 (MD) + 48 (internal) + userdata = total bytes per message |
In test above, 372+250 = 622, which divided into 4057 (usable space per page) = 6.5/page, or 6 rounded down. We are seeing that now.
This still doesn't explain the issue before(default pageset settings = 1000 primary/500 secondary). The only thing different is adding the parameter to the FORMAT command. The manual says the default is formatting 1 page. This may explain why we should not use the default (without the PAGES parameter).
Hope this helps others bypass this gotcha.  |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
