Subject | Re: [IB-Architect] New feature request (thread priority) |
---|---|
Author | Ann W. Harrison |
Post date | 2000-11-25T16:58:04Z |
At 01:16 PM 11/25/2000 +0100, Toni Martir wrote:
low priority threads blocking high priority threads and unable to
run because of their priority.
Assuming that problem could be solved - as Jim suggested by causing
the priority of a blocking transaction to be boosted - there are some
other questions.
Jason suggested allowing the priority to be adjusted on a statement
by statement basis. My immediate thought is that the transaction
parameter would be easier to force through intermediate layers other
than IBO, which is infinitely malleable. Nor is there any reason
why the capability should not be introduced at the transaction level
then added to the statement level.
Then there is my normal quibble - do we have enough flexibility in
the mechanism? Many years ago a very good programmer told me that
when you've used 4 bits in your flag byte, it's time to make it a
flag word. How about something more like: tpb_priority <n> where
<n> ranges from 0 - 99?
easy to implement, particularly in light of the blocking issues.
Regards,
Ann
www.ibphoenix.com
We have answers.
>The feature request is to create a new transaction parameter option like:Jim has just explained the first issue I have with this proposal -
>
>tpb_high_priority
>tpb_low_priority
low priority threads blocking high priority threads and unable to
run because of their priority.
Assuming that problem could be solved - as Jim suggested by causing
the priority of a blocking transaction to be boosted - there are some
other questions.
Jason suggested allowing the priority to be adjusted on a statement
by statement basis. My immediate thought is that the transaction
parameter would be easier to force through intermediate layers other
than IBO, which is infinitely malleable. Nor is there any reason
why the capability should not be introduced at the transaction level
then added to the statement level.
Then there is my normal quibble - do we have enough flexibility in
the mechanism? Many years ago a very good programmer told me that
when you've used 4 bits in your flag byte, it's time to make it a
flag word. How about something more like: tpb_priority <n> where
<n> ranges from 0 - 99?
>I think it's easy to implement (I'm studying Interbase source code now)My final quibble for the moment. I'd be quite surprised if it were
easy to implement, particularly in light of the blocking issues.
Regards,
Ann
www.ibphoenix.com
We have answers.