[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [xmlblaster] Durable QoS performance



> If you have 1000 different sensors, and every sensor has its 
> own ID (key 
> oid)
> this should be no problem at all.
> The first time a sensor publishes a message, it takes some 
> milliseconds
> more to initialize the message, all follower messages of this
> sensor (with the same oid) are processed much faster.
> 
> Note that for a new subscriber coming, the MoM only holds the newest
> value to deliver.

Which is reasonable.

> Subscribed clients have a callback queue to deliver all 
> messages in sequence
> (if no message priority is specified), so you have a virtual message 
> history here
> if a consuming client is slow in processing or unreachable 
> for some time.

What happens if you have a sequence of durable, undelivered messages on the
same oid and a subscriber is unreachable? The oid = persistence key seems to
indicate that there's only one persisted message with a given oid in the
system at one time. What I'm thinking of is the sequence

Subscriber Subscribes
Sensor sends message 1
Sensor sends message 2 [Can this happen before message 1 processed?]
Sensor sends message 3
Subscriber processes message 1
Subscriber processes message 2
Subscriber processes message 3

>                                               .... // we specify a 
> specific callback!!!

Don't mind me, burbling in the corner!

In any case, I can't reproduce the odd behaviour anymore, so it may be the
result of a inadvertant bug that I've somehow fixed without noticing. So all
is good.

-- 
Doug.Palmer at csiro.au
CSIRO Mathematical and Information Sciences
GPO Box 664                                    tel: +61 2 6216 7061
Canberra ACT 2601, Australia                   fax: +61 2 6216 7111