« Simple queue service » : différence entre les versions

De Banane Atomic
Aller à la navigationAller à la recherche
 
(4 versions intermédiaires par le même utilisateur non affichées)
Ligne 4 : Ligne 4 :


= Description =
= Description =
Message queue service, used to store messages while waiting for an application to process them.
Message queue service, used to store messages while waiting for an application to process them.<br>
Pull-based service.
It allows to decouple your infrastructure by acting as a buffer between your components.


= Workflow =
= Workflow =
Ligne 12 : Ligne 12 :
# if the message stays invisible longer than the visibility timeout it becomes visible again
# if the message stays invisible longer than the visibility timeout it becomes visible again
# when a application has completely processed a message, it is remove from the queue
# when a application has completely processed a message, it is remove from the queue
# if the message has not been proceeded after the retention period (1m to 14d), it could be moved to a dead letter queue


= Features =
= Features =
* Pull-based service
* Message size: 256 KB
* Message size: 256 KB
* Short polling
* Short polling

Dernière version du 26 avril 2024 à 15:05

Links

Description

Message queue service, used to store messages while waiting for an application to process them.
It allows to decouple your infrastructure by acting as a buffer between your components.

Workflow

  1. a message is picked up by an application
  2. the message is marked as invisible so that no other application can start processing that message
  3. if the message stays invisible longer than the visibility timeout it becomes visible again
  4. when a application has completely processed a message, it is remove from the queue
  5. if the message has not been proceeded after the retention period (1m to 14d), it could be moved to a dead letter queue

Features

  • Pull-based service
  • Message size: 256 KB
  • Short polling
    • API call to get 10 messages max
  • Long polling
    • API call and wait X seconds for incoming messages
    • less API calls: more efficient and cost effective
  • Standard
    • Order is NOT guaranteed
    • Messages delivered more than once (duplicates)
    • Great for performances
  • FIFO
    • Order is guaranteed
    • Messages only delivered once
    • Throughput limit 300 messages /s w/o batching, 3000 messages /s with batching