Queue-Primarily based Messaging in Home windows Azure

A typical messaging answer exchanges information between its distributed elements utilizing message queues, which embody publishers publishing messages into queues and subscribers supposed to obtain messages. The subscriber might be applied both as a single or a multiple-threaded course of, both constantly working or initiated on demand.

Main queuing mechanisms

At a better stage, there are two major queuing mechanisms used to allow a queue listener (receiver) to obtain messages saved in a queue.

1. Polling or Ballot-based mannequin: A listener displays a queue by checking the queue at common intervals. A listener is part of a employee function occasion. The primary processing logic is comprised of a loop wherein messages are dequeued and dispatched for processing. The listener checks for messages periodically. The queue is polled till the listener is notified to exit the loop. The Home windows Azure pricing mannequin measures storage transactions primarily based on requests carried out towards the queue, no matter whether or not the queue is empty or not.
2. Triggering or Push-based mannequin: A listener subscribes to an occasion triggered both by the writer or by the queue service supervisor, each time a message arrives on a queue. Then the listener dispatches the message for processing. So, it doesn’t must ballot the queue to find out whether or not any new work is offered or not. A notification might be pushed to the queue listeners for each new message, when the primary message arrives in an empty queue, or when the queue reaches a sure stage. Whereas utilizing Home windows Azure, the Service Bus quantity of messaging entities like queues or matters ought to be thought of.

Finest practices for optimizing transaction prices

In a queue-based messaging answer, the amount of storage transactions might be lowered utilizing a mix of the next strategies.

1. Group associated messages right into a single bigger batch, and compress and retailer the compressed picture within the blob storage, whereas holding a reference of the blob within the queue.
2. Batch a number of messages collectively in a single storage transaction. TheGetMessagesmethod within the Queue Service API allows de-queuing the desired variety of messages in a single transaction.
3. Whereas polling, keep away from aggressive polling intervals and implement a back-off delay that will increase the time between polling requests if a queue stays constantly empty.
4. Scale back the variety of queue listeners- when utilizing a pull-based mannequin, use just one queue listener per function occasion when a queue is empty. To additional cut back the variety of queue listeners per function occasion to zero, use a notification mechanism to instantiate queue listeners when the queue receives work objects.
5. If queues stay empty for more often than not, routinely scale down the variety of function cases and proceed to observe related system metrics to find out if and when the applying ought to scale up the variety of function cases to deal with growing workload.
6. Utilizing a mix of polling and push-based notifications, allows the listeners to subscribe to a notification occasion (set off) that’s raised upon sure situations to point {that a} new workload is placed on the queue.