Hierarchical dynamic scheduling
Selective resource migration is disclosed. A computer system includes physical memory and a plurality of physical processors. Each of the processors has one or more cores and each core instantiates one or more virtual processors that executes program code. Each core is configured to invoke a hyper-kernel on its hosting physical processor when the core cannot access a portion of the physical memory needed by the core. The hyper-kernel selectively moves the needed memory closer to a location accessible by the physical processor or remaps the virtual processor to another core.
1. A computer system, comprising:
a plurality of interconnected computing nodes, wherein a guest operating system is run on a virtual environment that is defined by a set of hyper-kernels running on the plurality of interconnected computing nodes, wherein each computing node includes one or more physical processors, and wherein each physical processor has one or more cores, and wherein each core has one or more hyperthreads;
wherein in response to receiving an indication that a core or a hyperthread should be assigned, a hyper-kernel determines, based at least in part on a search of a hierarchical data structure comprising one or more continuations, a continuation that is ready to run to assign to the core or the hyperthread to resume, wherein the continuation comprises a representation of a state of a virtual processor; and
wherein the continuation to assign to the core or the hyperthread is determined by the hyper-kernel at least in part by evaluating one or more cost functions.
2. The computer system recited in claim 1 wherein the search begins at a leaf level of the hierarchical data structure.
3. The computer system recited in claim 2 wherein in response to a threshold specified by a cost function being exceeded, the search is performed up the hierarchical data structure.
4. The computer system recited in claim 2 wherein the hierarchical data structure includes a plurality of work queues, wherein each work queue includes a set of continuations, and wherein in response to determining that a work queue at the leaf level is empty, the search is performed up the hierarchical data structure.
5. The computer system recited in claim 1 wherein the continuation is included in a scheduler object in the hierarchical data structure, and wherein evaluating the cost function comprises determining a height of the scheduler object in the hierarchical data structure.
6. The computer system recited in claim 1 wherein evaluating the cost function comprises determining an age of the continuation.
7. The computer system recited in claim 1 wherein evaluating the cost function comprises determining a last physical processor to run the continuation.
8. The computer system recited in claim 1 wherein evaluating the cost function comprises determining a last virtual processor to run the continuation.
9. The computer system recited in claim 1 wherein evaluating the cost function comprises determining a computing node on which the continuation was last executing.
10. The computer system recited in claim 1 wherein evaluating the cost function comprises determining group membership of the continuation.
11. A method, comprising:
receiving an indication that a core or a hyperthread should be assigned, wherein the core or the hyperthread is included in a physical processor, wherein the physical processor is included in a computing node, wherein the computing node is included in a plurality of physically interconnected computing nodes, and wherein a guest operating system is run on a virtual environment that is defined by a set of hyper-kernels running on the plurality of interconnected computing nodes; and
based at least in part on a search of a hierarchical data structure comprising one or more continuations, determining, by a hyper-kernel, a continuation that is ready to run to assign to the core or the hyperthread to resume, wherein the continuation comprises a representation of a state of a virtual processor; and
wherein the continuation to assign to the core or the hyperthread is determined by the hyper-kernel at least in part by evaluating one or more cost functions.
12. The method of claim 11 wherein the search begins at a leaf level of the hierarchical data structure.
13. The method of claim 12 wherein in response to a threshold specified by a cost function being exceeded, the search is performed up the hierarchical data structure.
14. The method of claim 12 wherein the hierarchical data structure includes a plurality of work queues, wherein each work queue includes a set of continuations, and wherein in response to determining that a work queue at the leaf level is empty, the search is performed up the hierarchical data structure.
15. The method of claim 11 wherein the continuation is included in a scheduler object in the hierarchical data structure, and wherein evaluating the cost function comprises determining a height of the scheduler object in the hierarchical data structure.
16. The method of claim 11 wherein evaluating the cost function comprises determining an age of the continuation.
17. The method of claim 11 wherein evaluating the cost function comprises determining a last physical processor to run the continuation.
18. The method of claim 11 wherein evaluating the cost function comprises determining a last virtual processor to run the continuation.
19. The method of claim 11 wherein evaluating the cost function comprises determining a computing node on which the continuation was last executing.
20. The method of claim 11 wherein evaluating the cost function comprises determining group membership of the continuation.