Throttling Concepts
When you use throttling to control message flow, a throttling queue is created in which messages are enqueued when a business service reaches its maximum concurrency or when a throttling group reaches its maximum concurrency. Messages with a higher priority are processed first. If messages have the same priority, they are processed on a first-in first-out basis. To ensure messages with a higher priority are processed first, assign priorities to messages using the routing options. The greater the integer for priority, the higher the priority is for the message.
A throttling queue is an in-memory queue. There is at most one queue per throttled business service or per server. Messages that are placed in this queue are not recoverable when a server fails or when you restart a server. Once a message has been in the throttling queue for an interval greater than the value of message expiration configured for the business service or throttling group, it becomes an expired message and is removed from the queue. When you delete or rename a business service, all the messages in the throttling queue are discarded.
Throttling Properties:
When you enable throttling for a business service or throttling group, you configure information about the message capacity and about the throttling queue used to hold messages when capacity is reached. To use a throttling queue, you must specify the queue length. If the throttling queue length is 0 (zero), messages are discarded once the defined capacity (maximum concurrency) is reached. When messages are discarded or removed from the queue due to exceeding the queue length or expiring, the throttling engine throws a back to the pipeline, and the error handler is triggered if one is configured.
Maximum Concurrency:
The maximum concurrency restricts the number of records that can be concurrently processed by a business service or throttling group. When this threshold is reached for a business service, all the incoming messages for the business service are placed in a throttling queue until the business service can accept more messages. If the queue is full, messages in the queue with a lower priority are removed and the new incoming messages are enqueued. Any change to this setting during runtime affects both new messages and those already in the queue. When you increase the value, Service Bus allows more messages to be processed once the messages in the queue are processed. When you decrease the value, Service Bus places any new messages in a throttling queue until the number of messages being processed goes below the new threshold.
Throttling Queue Length:
The throttling queue length limits the number of messages that can be held in the throttling queue at any given time. All incoming messages beyond the maximum concurrency limit are placed in the throttling queue. When the queue is full, the message in the queue with the lowest priority is removed if a new incoming message has a higher priority. When you decrease the value for this setting during runtime, all the messages beyond the new length are discarded.
Message Expiration (TTL):
The message expiration period (or time to live) limits the amount of time a message can stay in the throttling queue. When the time period has elapsed, the message is removed from the queue. These messages are referred to as expired messages. When you increase the value for this setting, the expiration time for the new messages and the messages that are already present in the queue is increased. When you decrease the value, all the messages that have exceeded the new value are immediately discarded. This value is set at the queue level; all messages put in the queue have the same Message Expiration value. When the time period is exceeded, messages are removed from the queue without being processed.
Throttling and Work Managers:
While endpoint throttling and dispatch policies (Work Managers) both work to limit loads, they work on different areas of processing. The Work Manager configured for a proxy service limits the number of threads running on that proxy service. A Work Manager configured on a business service limits the number of threads processing responses from the back-end system. Endpoint throttling configured on a business service limits requests to the back-end system. Using a combination of Work Managers and throttling gives you control over these three processing points.
Comments
Post a Comment