What is a messaging unit?
A messaging unit is a set of dedicated resources exclusively reserved for premium namespaces. This resource set can deliver consistent and repeatable performance of messaging workloads. Each premium namespace can have 1, 2 or 4 messaging units and the resource allocation grows linearly: 2 messaging units will consist of twice as many resources allocated as 1 messaging unit.
Related questions and answers
Relay hours are billed for the cumulative amount of time during which each Service Bus Relay is "open". A relay is implicitly instantiated and opened at a given Service Bus address (service namespace URL) when a relay-enabled WCF service, or "relay listener", first connects to that address. It's only closed when the last listener disconnects from its address. Therefore, for billing purposes a relay is considered "open" from the time the first relay listener connects until the time the last relay listener disconnects from the Service Bus address of that relay.
A hybrid connection allows you to establish bi-directional, binary stream communication between two networked applications, and one or both parties can reside behind NATs or firewalls. The listener that accepts this relayed connection and the sender that initiates the connection can both be implemented on any platform, and in any language, that has a basic WebSocket capability, including the WebSocket API in most web browsers.
The premium tier of Service Bus messaging is a flat daily rate per messaging unit purchased. Namespaces created as premium can have 1, 2 or 4 messaging units, which will each accrue the given number of messaging unit daily rate charges. Premium namespaces can have the number of purchased messaging units changed at any time, but the daily rate is based on the maximum number of message units assigned to the namespace at any time.
The premium tier of Service Bus messaging provides all the messaging features of Azure Service Bus queues and topics with predictable, repeatable performance, higher throughput and improved availability. The premium tier uses a dedicated resource allocation model to provide workload isolation and consistent performance. Because the compute and memory resources in the premium tier are dedicated, there are no per-message transaction charges as in other tiers. All transactions are included in the message unit allocation.
A brokered connection is defined as one of the following:
- An AMQP connection from a client into a Service Bus topic, subscription, queue or event hub.
- An HTTP call to receive a message from a Service Bus topic or queue that has a receive timeout value greater than zero. Microsoft charges for the peak number of concurrent brokered connections that exceed the included quantity (1,000 in the standard and premium tiers). Peaks are measured on an hourly basis, prorated by dividing by 730 hours in a month and added up over the monthly billing period. The included quantity (1,000 brokered connections per month) is applied at the end of the billing period against the sum of the prorated hourly peaks. Examples:
- 5,000 clients each connect via a single AMQP connection and receive commands from a Service Bus topic and send events to queues. If all clients connect for 12 hours every day, you will see the following connection charges (in addition to any other Service Bus charges): 5,000 connections * 12 hours * 30.5 days/730 = 2,500 brokered connections. After the monthly allowance of 1,000 brokered connections, you would be charged for 1,500 brokered connections.
- 5,000 clients receive messages from a Service Bus queue via HTTP, specifying a non-zero timeout. If all devices connect for 12 hours every day, you will see the following connection charges (in addition to any other Service Bus charges): 5,000 HTTP receive connections * 12 hours per day * 30.5 days/730 hours = 2,500 brokered connections.
The relay counts each message sent to the relay, and each message sent by the relay, as billable. A billable message is a data frame of at most 64 KB. If a message exceeds 64 KB, such as an HTTP reply that returns an image, each further 64 KB counts as an additional billable message. For a normal relayed service that implements a request/response scheme, the request first travels to the relay, then to the service, and the reply traverses the same path. That amounts to at least four billable messages. For a multicast service that has 4 listeners, the message sent to the relay counts as 1 message, and the 4 messages sent to the listeners also each count as a message, resulting in a total of 5 messages.