Process Scheduling to Maximize Input Throughput
A schedule graph may be used to identify executable elements that consume data from a network interface or other input/output interface. The schedule graph may be traversed to identify a sequence or pipeline of executable elements that may be triggered from data received on the interface, then a process scheduler may cause those executable elements to be executed on available processors. A queue manager and a load manager may optimize the resources allocated to the executable elements to maximize the throughput for the input/output interface. Such as system may optimize processing for input or output of network connections, storage devices, or other input/output devices.
1 . A system comprising:
a plurality of processors;
an operating system executing on said plurality of processors;
a queue manager that:
receives a scheduling graph comprising executable elements and relationships between said executable elements, said relationships comprising data dependency relationships;
from said scheduling graph, identifies a set of executable elements dependent on a first executable element that receives a first input; and
schedules said set of executable elements such that a process scheduler causes said set of executable elements to process said first input.
2 . The system of claim 1 , said queue manager that further:
from said scheduling graph, identifies a second set of executable elements from which said first executable element is dependent; and
schedules said second set of executable elements.
3 . The system of claim 2 , said second set of executable elements being executed at an elevated priority.
4 . The system of claim 1 further comprising:
a load manager that:
identifies a second executable element downstream from said first executable element, said second executable element being a chokepoint for processing output from said first executable element; and
allocates additional processing resources to said second executable element.
5 . The system of claim 1 , said source being a network input device.
6 . The system of claim 1 , said source being a data storage device.
7 . A method comprising:
receiving a scheduling graph comprising executable elements and relationships between said executable elements;
identifying a first executable element within said scheduling graph;
identifying a second executable element from said scheduling graph, said second executable element having a data dependency relationship with said first executable element;
scheduling said second executable element such that a process scheduler causes said second executable element to be executed after said first executable element.
8 . The method of claim 7 , said second executable element not having a message passing relationship with said first executable element.
9 . The method of claim 8 further comprising:
identifying a third executable element from said scheduling graph, said first executable element being dependent on said third executable element; and
scheduling said third executable element such that said process scheduler causes said third executable element to be executed prior to said first executable element.
10 . The method of claim 7 , said second executable element being scheduled with an elevated priority.
11 . The method of claim 7 further comprising:
receiving a plurality of input data items consumed by said first executable element.
12 . The method of claim 11 further comprising:
identifying a fourth executable element being dependent on said first executable element, said fourth executable element being causing said first executable element to wait to process said plurality of input data items; and
allocating additional memory resources to said fourth executable element.
13 . The method of claim 12 further comprising:
determining a current rate at which said first executable element processes said input data items; and
measuring a change in said current rate when said fourth executable element has been allocated said additional processing resources.
14 . The method of claim 13 further comprising:
when said change is positive, adding additional memory resources; and
when said change is not positive, removing at least some of said additional memory resources.
15 . The method of claim 14 further comprising:
updating said scheduling graph with said additional memory resources.