View previous topic :: View next topic |
Author |
Message |
Bibhakar Beginner
Joined: 04 Aug 2005 Posts: 3 Topics: 1
|
Posted: Thu Aug 04, 2005 12:58 pm Post subject: Generation of Unique Ids |
|
|
Hi - We are designing a system that will be distributed across 32 processors on distributed side and 16 on M/f side. This application will be accessed by possibly 1200 conconcurrent users. We are trying capture unique event record for each action that happens in the application. (kind of audit trail). We are looking at generating an unique id w/o getting back to the tables because of heaviness of transactions.We can't depend on T/S as we still can have duplicates w.r.t that - Any insight is really appriciated. |
|
Back to top |
|
 |
kolusu Site Admin

Joined: 26 Nov 2002 Posts: 12378 Topics: 75 Location: San Jose
|
Posted: Thu Aug 04, 2005 1:27 pm Post subject: |
|
|
Bibhakar,
what type of transactions are we talking here? CICS ? IMS? JAVA?
kolusu _________________ Kolusu
www.linkedin.com/in/kolusu |
|
Back to top |
|
 |
Bibhakar Beginner
Joined: 04 Aug 2005 Posts: 3 Topics: 1
|
Posted: Thu Aug 04, 2005 1:35 pm Post subject: |
|
|
CICS, MVS and Java Back-end DB2 |
|
Back to top |
|
 |
Bibhakar Beginner
Joined: 04 Aug 2005 Posts: 3 Topics: 1
|
Posted: Thu Aug 04, 2005 1:44 pm Post subject: |
|
|
And we have to generate these unique id in CICS, MVS as well as in Java w/o going back to DB2 for verification.. |
|
Back to top |
|
 |
semigeezer Supermod
Joined: 03 Jan 2003 Posts: 1014 Topics: 13 Location: Atlantis
|
Posted: Thu Aug 04, 2005 6:19 pm Post subject: |
|
|
Are these IDs to be permanent? In other words, can they be reused, or do you just want to verify that no one else in the system has that id at any given time? If they can be reused, you might be able to use a small long running program on MVS that just sets up an ENQ with a given unique name. Or for even less resource usage, the program could keep a list of used and unused ids (any of a million implentations are possible, but the fastest would be a binary tree of that you search and change as needed). You'd also need to write some mechanism of communicating with it, presumably via TCPIP sockets, but if it is in assembler or any reasonably efficient language, it would be very fast. There are, of course problems like what happens if a process dies and never unregisters, but I guess you'd have that even if you were using something big like DB2 as a back end. I'm not sure I understand the problem, but that never stopped me from giving an answer  |
|
Back to top |
|
 |
|
|