MVSFORUMS.com

[deepa12]

1.1 When a GU is issued on a child segment, is the entire DB record locked? If so is it an exclusive lock or shared lock?


3. What is the logic to handle i/p messges from different terminals from i/p Q. I understand the Message I/p queue has i/ps requests for the same transaction queued from many terminals 1 after another. But in Telon-IMS or anywhere normally i have not seen logic which checks LTERM & ensures that a new process for another terminal has to start. Is it implied by some status code. I only know that QD means end of messages.

Can someone elaborate on this logic. CICS transactions does not have the message Q concept and it handles data from 1 terminal at a time

4. Does IMS make physical updates to the db even before commit point & rollback changes when say for e.g. the txn crashes in between & commit has not reached?

I think it has a write ahead buffer in Virtual storage for uncommitted transaction and a system log for automatic system restart against crashes. It has 1 more log for transaction recovery which it maintains for committed transaction Pls correct my understanding.

4.2 In IMS is there a rollback transaction like in db2

4.3 I also understand that the concept of commit , log maintanence is different for fastpath MPP. Can you pls elaborate on this by comparing the MPP

5. How to distinguish between fastpath MPP & normal MPP

6. In IMS-TELON what is special about direct switching
_________________
deepa

[Bithead]

Deepa12,

I do not use Fast Path, nor am I familiar with TELON, but I will attempt to cover some of the answers to your questions and guide you to where you might find more information.

1.1 - The lock type depends on the intent (from PROCOPT and call type) and the physical database dataset type. Typically it affects blocks rather than segments and can be physical or logical. The Administrators Guide:Database Manager has a more in depth explanation.

3. - Each GU to the message queue gets a new message. This could be from the same or another source. If the transaction is single-threaded and you have 50 messages, it will process one at a time until all are processed or until the scheduling alogorithm determines that another transaction should take it's place. If the transaction is configured to run in more than one region at a time, then the queue will be split between the two, although it might not be an even split.

4. - Updates are typically written to a buffer in IMS TM rather than to the physical data. Actual update occurs when a checkpoint (explicit or implicit) is taken or when the buffer is needed for something else. IMS initially writes log records to the WADS and then these get moved to the OLDS. There are a number of each of these log dataset types to allow for logs to be cycled. The OLDS should also have a primary and secondary to allow for recovery in the event of disk failure. When the OLDS is filled, IMS switches to the next OLDS and then archives the data, typically to tape.

4.2 - There are rollback commands - see the programming guide.

[deepa12]

Hi..
Thanks for the replies..Ya.. will look at the APPLICATION PGMERS GUIDE & DB ADMIN GUIDE
Can someone else pls help me with the rest of my doubts?
_________________
deepa