摘要:
Systems and methods are provided for auto-balancing of throughput in a real-time event-driven system. The systems and methods may include identifying a plurality of processing stages for a real-time event-driven system that receives a system-level event and provides a system-level output, wherein processing stages between a first processing stage and a last processing stage are arranged in a sequential order such that a respective output of an earlier processing stage is provided to a respective request queue of a later processing stage; determining a respective number of requests in each request queue of the plurality of processing stages; comparing the respective determined number of requests to at least one threshold to determine that at least one processing stage is over-utilized, and for the at least one processing stage that is determined to be over-utilized, increasing an allocation of hardware resources to the respective over-utilized processing stage.
摘要:
Systems and methods may include providing a first silo configured to operate a first set of event-driven processing stages, and a second silo configured to operate a second set of event-driven processing stages; arranging at least one shared transaction cache between the first set of processing stages and the second set of processing stages of the respective first and second silos, where each processing stage of the first and second sets of processing stages can obtain or receive one or more transactions from the at least one shared transaction cache for processing, and where when a distressed processing stage in the first silo fails to process a first transaction designated for the distressed processing stage in a predefined amount of time, a non-distressed processing stage in the second silo obtains or receives the first transaction from the at least one shared transaction cache to process the first transaction on behalf of the distressed processing stage.