Storage queues and Service Bus queues - compared and contrasted
This article analyzes the differences and similarities between the two types of queues offered by Microsoft Azure: Storage queues and Service Bus queues. By using this information, you can make a more informed decision about which solution best meets your needs.
Introduction
Azure supports two types of queue mechanisms: Storage queues and Service Bus queues.
Storage queues are part of the Azure Storage infrastructure. They allow you to store large numbers of messages. You access messages from anywhere in the world via authenticated calls using HTTP or HTTPS. A queue message can be up to 64 KB in size. A queue might contain millions of messages, up to the total capacity limit of a storage account. Queues are commonly used to create a backlog of work to process asynchronously. For more information, see What are Azure Storage queues.
Service Bus queues are part of a broader Azure messaging infrastructure that supports queuing, publish/subscribe, and more advanced integration patterns. They're designed to integrate applications or application components that might span multiple communication protocols, data contracts, trust domains, or network environments. For more information about Service Bus queues/topics/subscriptions, see the Service Bus queues, topics, and subscriptions.
Technology selection considerations
Storage queues and Service Bus queues have a slightly different feature set. You can choose either one or both, depending on the needs of your particular solution.
When determining which queuing technology fits the purpose of a given solution, solution architects and developers should consider these recommendations.
Consider using Storage queues
As a solution architect/developer, you should consider using Storage queues when:
- Your application must store over 80 gigabytes of messages in a queue.
- Your application wants to track progress for processing a message in the queue. It's useful if the worker processing a message crashes. Another worker can then use that information to continue from where the prior worker left off.
- You require server side logs of all of the transactions executed against your queues.
Consider using Service Bus queues
As a solution architect/developer, you should consider using Service Bus queues when:
- Your solution needs to receive messages without having to poll the queue. With Service Bus, you can achieve it by using a long-polling receive operation using the TCP-based protocols that Service Bus supports.
- Your solution requires the queue to provide a guaranteed first-in-first-out (FIFO) ordered delivery.
- Your solution needs to support automatic duplicate detection.
- You want your application to process messages as parallel long-running streams (messages are associated with a stream using the session ID property on the message). In this model, each node in the consuming application competes for streams, as opposed to messages. When a stream is given to a consuming node, the node can examine the state of the application stream state using transactions.
- Your solution requires transactional behavior and atomicity when sending or receiving multiple messages from a queue.
- Your application handles messages that can exceed 64 KB but won't likely approach the 256 KB or 1 MB limit, depending on the chosen service tier (although Service Bus queues can handle messages up to 100 MB).
- You deal with a requirement to provide a role-based access model to the queues, and different rights/permissions for senders and receivers. For more information, see the following articles:
- Your queue size won't grow larger than 80 GB.
- You want to use the AMQP 1.0 standards-based messaging protocol. For more information about AMQP, see Service Bus AMQP Overview.
- You envision an eventual migration from queue-based point-to-point communication to a publish-subscribe messaging pattern. This pattern enables integration of additional receivers (subscribers). Each receiver receives independent copies of either some or all messages sent to the queue.
- Your messaging solution needs to support the "At-Most-Once" and the "At-Least-Once" delivery guarantees without the need for you to build the additional infrastructure components.
- Your solution needs to publish and consume batches of messages.
Compare Storage queues and Service Bus queues
The tables in the following sections provide a logical grouping of queue features. They let you compare, at a glance, the capabilities available in both Azure Storage queues and Service Bus queues.
Foundational capabilities
This section compares some of the fundamental queuing capabilities provided by Storage queues and Service Bus queues.
Comparison Criteria | Storage queues | Service Bus queues |
---|---|---|
Ordering guarantee | No For more information, see the first note in the Additional Information section. |
Yes - First-In-First-Out (FIFO) (by using message sessions) |
Delivery guarantee | At-Least-Once | At-Least-Once (using PeekLock receive mode. It's the default) At-Most-Once (using ReceiveAndDelete receive mode) Learn more about various Receive modes |
Atomic operation support | No | Yes |
Receive behavior | Non-blocking (completes immediately if no new message is found) |
Blocking with or without a timeout (offers long polling, or the "Comet technique") Non-blocking (using .NET managed API only) |
Push-style API | No | Yes Our .NET, Java, JavaScript, and Go SDKs provide push-style API. |
Receive mode | Peek & Lease | Peek & Lock Receive & Delete |
Exclusive access mode | Lease-based | Lock-based |
Lease/Lock duration | 30 seconds (default) 7 days (maximum) (You can renew or release a message lease using the UpdateMessage API.) |
30 seconds (default) You can renew the message lock for the same lock duration each time manually or use the automatic lock renewal feature where the client manages lock renewal for you. |
Lease/Lock precision | Message level Each message can have a different timeout value, which you can then update as needed while processing the message, by using the UpdateMessage API. |
Queue level (each queue has a lock precision applied to all of its messages, but the lock can be renewed as described in the previous row) |
Batched receive | Yes (explicitly specifying message count when retrieving messages, up to a maximum of 32 messages) |
Yes (implicitly enabling a pre-fetch property or explicitly by using transactions) |
Batched send | No | Yes (by using transactions or client-side batching) |
Additional information
- Messages in Storage queues are typically first-in-first-out, but sometimes they can be out of order. For example, when the visibility-timeout duration of a message expires because a client application crashed while processing a message. When the visibility timeout expires, the message becomes visible again on the queue for another worker to dequeue it. At that point, the newly visible message might be placed in the queue to be dequeued again.
- The guaranteed FIFO pattern in Service Bus queues requires the use of messaging sessions. If the application crashes while it's processing a message received in the Peek & Lock mode, the next time a queue receiver accepts a messaging session, it will start with the failed message after the session's lock duration expires.
- Storage queues are designed to support standard queuing scenarios, such as the following ones:
- Decoupling application components to increase scalability and tolerance for failures
- Load leveling
- Building process workflows.
- Inconsistencies regarding message handling in the context of Service Bus sessions can be avoided by using session state to store the application's state relative to the progress of handling the session's message sequence, and by using transactions around settling received messages and updating the session state. This kind of consistency feature is sometimes labeled exactly once processing in other vendor's products. Any transaction failures will obviously cause messages to be redelivered and that's why the term isn't exactly adequate.
- Storage queues provide a uniform and consistent programming model across queues, tables, and BLOBs – both for developers and for operations teams.
- Service Bus queues provide support for local transactions in the context of a single queue.
- The Receive and Delete mode supported by Service Bus provides the ability to reduce the messaging operation count (and associated cost) in exchange for lowered delivery assurance.
- Storage queues provide leases with the ability to extend the leases for messages. This feature allows the worker processes to maintain short leases on messages. So, if a worker crashes, the message can be quickly processed again by another worker. Also, a worker can extend the lease on a message if it needs to process it longer than the current lease time.
- Storage queues offer a visibility timeout that you can set upon the enqueuing or dequeuing of a message. Also, you can update a message with different lease values at run-time, and update different values across messages in the same queue. Service Bus lock timeouts are defined in the queue metadata. However, you can renew the message lock for the pre-defined lock duration manually or use the automatic lock renewal feature where the client manages lock renewal for you.
- The maximum timeout for a blocking receive operation in Service Bus queues is 24 days. However, REST-based timeouts have a maximum value of 55 seconds.
- Client-side batching provided by Service Bus enables a queue client to batch multiple messages into a single send operation. Batching is only available for asynchronous send operations.
- Features such as the 200-TB ceiling of Storage queues (more when you virtualize accounts) and unlimited queues make it an ideal platform for SaaS providers.
- Storage queues provide a flexible and performant delegated access control mechanism.
Advanced capabilities
This section compares advanced capabilities provided by Storage queues and Service Bus queues.
Comparison Criteria | Storage queues | Service Bus queues |
---|---|---|
Scheduled delivery | Yes | Yes |
Automatic dead lettering | No | Yes |
Increasing queue time-to-live value | Yes (via in-place update of visibility timeout) |
Yes (provided via a dedicated API function) |
Poison message support | Yes | Yes |
In-place update | Yes | Yes |
Server-side transaction log | Yes | No |
Storage metrics | Yes Minute Metrics provides real-time metrics for availability, TPS, API call counts, error counts, and more. They're all in real time, aggregated per minute and reported within a few minutes from what just happened in production. For more information, see About Storage Analytics Metrics. |
Yes For information about metrics supported by Azure Service Bus, see Message metrics. |
State management | No | Yes (Active, Disabled, SendDisabled, ReceiveDisabled. For details on these states, see Queue status) |
Message autoforwarding | No | Yes |
Purge queue function | Yes | No |
Message groups | No | Yes (by using messaging sessions) |
Application state per message group | No | Yes |
Duplicate detection | No | Yes (configurable on the sender side) |
Browsing message groups | No | Yes |
Fetching message sessions by ID | No | Yes |
Additional information
- Both queuing technologies enable a message to be scheduled for delivery at a later time.
- Queue autoforwarding enables thousands of queues to autoforward their messages to a single queue, from which the receiving application consumes the message. You can use this mechanism to achieve security, control flow, and isolate storage between each message publisher.
- Storage queues provide support for updating message content. You can use this functionality for persisting state information and incremental progress updates into the message so that it can be processed from the last known checkpoint, instead of starting from scratch. With Service Bus queues, you can enable the same scenario by using message sessions. For more information, see Message session state.
- Service Bus queues support dead lettering. It can be useful for isolating messages that meet the following criteria:
- Messages can't be processed successfully by the receiving application
- Messages can't reach their destination because of an expired time-to-live (TTL) property. The TTL value specifies how long a message remains in the queue. With Service Bus, the message will be moved to a special queue called $DeadLetterQueue when the TTL period expires.
- To find "poison" messages in Storage queues, when dequeuing a message the application examines the DequeueCount property of the message. If DequeueCount is greater than a given threshold, the application moves the message to an application-defined "dead letter" queue.
- Storage queues enable you to obtain a detailed log of all of the transactions executed against the queue, and aggregated metrics. Both of these options are useful for debugging and understanding how your application uses Storage queues. They're also useful for performance-tuning your application and reducing the costs of using queues.
- Message sessions supported by Service Bus enable messages that belong to a logical group to be associated with a receiver. It creates a session-like affinity between messages and their respective receivers. You can enable this advanced functionality in Service Bus by setting the session ID property on a message. Receivers can then listen on a specific session ID and receive messages that share the specified session identifier.
- The duplication detection feature of Service Bus queues automatically removes duplicate messages sent to a queue or topic, based on the value of the message ID property.
Capacity and quotas
This section compares Storage queues and Service Bus queues from the perspective of capacity and quotas that might apply.
Comparison Criteria | Storage queues | Service Bus queues |
---|---|---|
Maximum queue size | 500 TB (limited to a single storage account capacity) |
1 GB to 80 GB (defined upon creation of a queue and enabling partitioning – see the “Additional Information” section) |
Maximum message size | 64 KB (48 KB when using Base 64 encoding) Azure supports large messages by combining queues and blobs – at which point you can enqueue up to 200 GB for a single item. |
256 KB or 100 MB (including both header and body, maximum header size: 64 KB). Depends on the service tier. |
Maximum message TTL | Infinite (api-version 2017-07-27 or later) | TimeSpan.MaxValue |
Maximum number of queues | Unlimited | 10,000 (per service namespace) |
Maximum number of concurrent clients | Unlimited | 5,000 |
Additional information
- Service Bus enforces queue size limits. The maximum queue size is specified when creating a queue. It can be between 1 GB and 80 GB. If the queue's size reaches this limit, additional incoming messages will be rejected and the caller receives an exception. For more information about quotas in Service Bus, see Service Bus Quotas.
- In the Standard messaging tier, you can create Service Bus queues and topics in 1 (default), 2, 3, 4, or 5-GB sizes. When enabling partitioning in the Standard tier, Service Bus creates 16 copies (16 partitions) of the entity, each of the same size specified. As such, if you create a queue that's 5 GB in size, with 16 partitions the maximum queue size becomes (5 * 16) = 80 GB. You can see the maximum size of your partitioned queue or topic in the Azure portal.
- With Storage queues, if the content of the message isn't XML-safe, then it must be Base64 encoded. If you Base64-encode the message, the user payload can be up to 48 KB, instead of 64 KB.
- With Service Bus queues, each message stored in a queue is composed of two parts: a header and a body. The total size of the message can't exceed the maximum message size supported by the service tier.
- When clients communicate with Service Bus queues over the TCP protocol, the maximum number of concurrent connections to a single Service Bus queue is limited to 100. This number is shared between senders and receivers. If this quota is reached, requests for additional connections will be rejected and an exception will be received by the calling code. This limit isn't imposed on clients connecting to the queues using REST-based API.
- If you require more than 10,000 queues in a single Service Bus namespace, you can contact the Azure support team and request an increase. To scale beyond 10,000 queues with Service Bus, you can also create additional namespaces using the Azure portal.
Management and operations
This section compares the management features provided by Storage queues and Service Bus queues.
Comparison Criteria | Storage queues | Service Bus queues |
---|---|---|
Management protocol | REST over HTTP/HTTPS | REST over HTTPS |
Runtime protocol | REST over HTTP/HTTPS | REST over HTTPS AMQP 1.0 Standard (TCP with TLS) |
.NET API | Yes (.NET Storage Client API) |
Yes (.NET Service Bus API) |
Native C++ | Yes | Yes |
Java API | Yes | Yes |
PHP API | Yes | Yes |
Node.js API | Yes | Yes |
Arbitrary metadata support | Yes | No |
Queue naming rules | Up to 63 characters long (Letters in a queue name must be lowercase.) |
Up to 260 characters long (Queue paths and names are case-insensitive.) |
Get queue length function | Yes (Approximate value if messages expire beyond the TTL without being deleted.) |
Yes (Exact, point-in-time value.) |
Peek function | Yes | Yes |
Additional information
- Storage queues provide support for arbitrary attributes that can be applied to the queue description, in the form of name/value pairs.
- Both queue technologies offer the ability to peek a message without having to lock it, which can be useful when implementing a queue explorer/browser tool.
- The Service Bus .NET brokered messaging APIs use full-duplex TCP connections for improved performance when compared to REST over HTTP, and they support the AMQP 1.0 standard protocol.
- Names of Storage queues can be 3-63 characters long, can contain lowercase letters, numbers, and hyphens. For more information, see Naming Queues and Metadata.
- Service Bus queue names can be up to 260 characters long and have less restrictive naming rules. Service Bus queue names can contain letters, numbers, periods, hyphens, and underscores.
Authentication and authorization
This section discusses the authentication and authorization features supported by Storage queues and Service Bus queues.
Comparison Criteria | Storage queues | Service Bus queues |
---|---|---|
Authentication | Symmetric key and Role-based access control (RBAC) | Symmetric key and Role-based access control (RBAC) |
Identity provider federation | Yes | Yes |
Additional information
- Every request to either of the queuing technologies must be authenticated. Public queues with anonymous access aren't supported.
- Using shared access signature (SAS) authentication, you can create a shared access authorization rule on a queue that can give users a write-only, read-only, or full access. For more information, see Azure Storage - SAS authentication and Azure Service Bus - SAS authentication.
- Both queues support authorizing access using Microsoft Entra ID. Authorizing users or applications using OAuth 2.0 token returned by Microsoft Entra ID provides superior security and ease of use over shared access signatures (SAS). With Microsoft Entra ID, there's no need to store the tokens in your code and risk potential security vulnerabilities. For more information, see Azure Storage - Microsoft Entra authentication and Azure Service Bus - Microsoft Entra authentication.
Conclusion
By gaining a deeper understanding of the two technologies, you can make a more informed decision on which queue technology to use, and when. The decision on when to use Storage queues or Service Bus queues clearly depends on many factors. These factors depend heavily on the individual needs of your application and its architecture.
You might prefer to choose Storage queues for reasons such as the following ones:
- If your application already uses the core capabilities of Microsoft Azure
- If you require basic communication and messaging between services
- Need queues that can be larger than 80 GB in size
Service Bus queues provide many advanced features such as the following ones. So, they might be a preferred choice if you're building a hybrid application or if your application otherwise requires these features.
- Sessions
- Transactions
- Duplicate detection
- Automatic dead-lettering
- Durable publish and subscribe capabilities
Next steps
The following articles provide more guidance and information about using Storage queues or Service Bus queues.
Feedback
https://aka.ms/ContentUserFeedback.
Disponibbli dalwaqt: Matul l-2024 se nkunu qed inwaqqfu gradwalment Problemi GitHub bħala l-mekkaniżmu ta’ feedback għall-kontenut u se nibdluh b’sistema ġdida ta’ feedback. Għal aktar informazzjoni, ara:Issottometti u ara feedback għal