Storage Queues vs Service Bus queues

Interesante la tabla comparativa que leo en http://preps2.wordpress.com/2011/09/17/comparison-of-windows-azure-storage-queues-and-service-bus-queues/ sobre las diferencias entre las colas de storage y las de service bus.

Aunque como se verá a continuación el sistema de colas de service bus es mucho más completo que el existente en el storage, no se debe caer en la tentación de emplear siempre el más completo, sino en aquel que sea mejor para la aplicación que se esté desarrollando.

Feature

Windows Azure Storage Queues

Service Bus Queues

Comments

Programming Models
Raw REST/HTTP Yes Yes
.NET API Yes(Windows Azure Managed Library) Yes(AppFabric SDK)
Windows Communication Foundation (WCF) binding No Yes
Windows Workflow Foundation (WF) integration No Yes
Protocols
Runtime REST over HTTP REST over HTTPBi-directional TCP The Service Bus managed API leverages the bi-directional TCP protocol for improved performance over REST/HTTP.
Management REST over HTTP REST over HTTP
Messaging Fundamentals
Ordering Guarantees No First-In-First-Out (FIFO) Note: guaranteed FIFO requires the use of sessions.
Message processing guarantees At-Least-Once (ALO) At Least-Once (ALO)Exactly-Once (EO) The Service Bus generally supports the ALO guarantee; however EO can be supported by using SessionState to store application state and using transactions to atomically receive messages and update the SessionState. The AppFabric workflow uses this technique to provide EO processing guarantees.
Peek Lock YesVisibility timeout: default=30s; max=2h YesLock timeout: default=30s; max=5m Windows Azure queues offer a visibility timeout to be set on each receive operation, while Service Bus lock timeouts are set per entity.
Duplicate Detection No Yes, send-side duplicate detection The Service Bus will remove duplicate messages sent to a queue/topic (based on MessageId).
Transactions No Partial The Service Bus supports local transactions involving a single entity (and its children). Transactions can also include updates to SessionState.
Receive Behavior Non-blocking, i.e., return immediately if no messages REST/HTTP: long poll based on user-provided timeout.NET API: 3 options: blocking, blocking with timeout, non-blocking.
Batch Receive Yes(explicit) Yes. Either (a) Implicitly using prefetch, or (b) explicitly using transactions.
Batch Send No Yes (using transactions)
Receive and Delete No Yes Ability to reduce operation count (and associated cost) in exchange for lowered delivery assurance.
Advanced Features
Dead lettering No Yes Windows Azure queues offer a ‘dequeue count’ on each message, so applications can choose to delete troublesome messages themselves.
Session Support No Yes Ability to have logical subgroups within a queue or topic.
Session State No Yes Ability to store arbitrary metadata with sessions. Required for integration with Workflow.
Message Deferral No Yes Ability for a receiver to defer a message until they are prepared to process it. Required for integration with Workflow.
Scheduled Delivery No Yes Allows a message to be scheduled for delivery at some future time.
Security
Authentication Windows Azure credentials ACS roles ACS allows for three distinct roles: admin, sender and receiver. Windows Azure has a single role with total access, and no ability for delegation.
Management Features
Get Message Count Approximate No Service Bus queues offer no operational insight at this point, but plan to in the future.
Clear Queue Yes No Convenience functions to clear queue efficiently.
Peek / Browse Yes No Windows Azure queues offer the ability to peek a message without locking it, which can be used to implement browse functionality.
Arbitrary Metadata Yes No Windows Azure queues allow an arbitrary set of <key, value> pairs on queue metadata.
Quotas/Limits
Maximum message size 8KB 256KB
Maximum queue size Unlimited 5GB Specified at queue creation, with specific values of 1,2,3,4 or 5 GB.
Maximum number of entities per service namespace n/a

10,000

Ibon Landa

bon Landa lleva más de 15 años dedicado al desarrollo de software. Durante este tiempo ha trabajado en diferentes empresas en las cuáles ha podido trabajar en diferentes entornos y tecnologías. Actualmente está focalizado principalmente en tareas de desarrollo, arquitectura, en las herramientas del ciclo de vida y en todo lo relacionado con la plataforma de Cloud Computing Microsoft Azure, área en el que ha sido reconocido como MVP. Participa de forma activa en la comunidad, escribiendo su blog, manteniendo un portal sobre Microsoft Azure y colaborando con Microsoft y grupos de usuarios en eventos de formación, talleres y giras de producto.

Deja un comentario

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *