ietf-smtp
[Top] [All Lists]

Re: Comments on: draft-melnikov-smtp-priority-02

2011-07-28 11:22:22

At 15:19 -0700 on 07/11/2011, Bill McQuillan wrote about Comments on: draft-melnikov-smtp-priority-02:

Section 5.1:

Since the vast majority of messages will either be marked as
"PRIORITY 0" or presumed to be priority 0, is there any mechanism
to prevent starvation for those messages marked at a negative
priority? Note that for a high volume MTA there would likely
always be SOME priority 0 messages in the queue which would
preclude sending ANY messages at a lower priority.

Many queuing systems that I have seen reward items that have a low priority and thus have sat on the queue and never processed [due to there always being higher priority items on the queue that prevent their being selected] by periodically bumping their priority. In this type of system, you would, for example, designate that a low priority item gets its priority bumped one priority notch after it has been unprocessed for x minutes (until it gets to priority 0/normal). Thus if the bump interval is 15 minutes, a priority -3 will go normal after no longer than 45 minutes (and thus be sent as it gets to the top of the priority 0 queue [which I assume will be FIFO within priority] either due to FIFO on queue or submission timestamp [which would make the delayed and bumped item the highest priority 0 item]).

<Prev in Thread] Current Thread [Next in Thread>