Virtual machine input/output thread management
A method performed by a physical computing system includes detecting an interrupt signal sent to a virtual processor being managed by the hypervisor, creating a map between the virtual processor and an Input/Output (I/O) thread associated with the interrupt signal, determining that the virtual processor is idle, finding the I/O thread associated with the idle virtual processor based on the map, and moving the I/O thread associated with the idle virtual processor up in a processing queue, the processing queue being for processes to be executed on a physical processor.
1. A method performed by a physical computing system, the method comprising:
with a hypervisor, detecting an interrupt signal sent to a virtual processor being managed by the hypervisor;
creating a map between the virtual processor and an Input/Output (I/O) thread associated with the interrupt signal; and
in response to determining that the virtual processor is idle;
finding the I/O thread associated with the idle virtual processor based on the map; and
moving the I/O thread associated with the idle virtual processor up in a processing queue, the processing queue being for processes to be executed on a physical processor.
2. The method of claim 1 , wherein the map comprises a data structure that matches I/O threads with corresponding virtual processors.
3. The method of claim 1 , wherein an association within the map is made between the virtual processor and the I/O thread if the I/O thread sends a threshold number of interrupt signals to the virtual processor.
4. The method of claim 1 , further comprising, storing the map in hypervisor memory.
5. The method of claim 1 , wherein the map associates the I/O thread with the virtual processor and a virtual machine corresponding to the virtual processor.
6. The method of claim 1 , wherein the virtual processor is idle because the virtual processor executed a halt command.
7. The method of claim 1 , wherein the moving the I/O thread associated with the idle virtual processor up in a processing queue comprises lending execution time from the virtual processor to the I/O thread.
8. The method of claim 1 , wherein the map includes associations of multiple I/O threads to multiple additional virtual processors.
9. The method of claim 1 , wherein the map uses a thread identifier of the I/O thread to map the I/O thread to the virtual processor.
10. A computing system comprising:
a processor; and
a memory comprising computer readable instructions that when executed by the processor, cause the system to:
run a hypervisor to host a virtual machine that includes a number of virtual processors;
run a scheduler to manage a processing queue for execution of threads on the processor;
detect an interrupt signal sent to one of the virtual processors;
create a map between an Input/Output (I/O) thread associated with the interrupt signal and the one of the virtual processors; and
determine that the virtual processor is in an idle state, and in response;
use the map to find the I/O thread associated with the virtual processor in the idle state; and
cause the scheduler to move the I/O thread up in the processing queue.
11. The system of claim 10 , wherein to create the map, the processor is further to cause the system to create an entry in a data structure, the entry associating the I/O thread with the one of the virtual processors.
12. The system of claim 11 , wherein the entry in the data structure includes a thread identifier for the I/O thread.
13. The system of claim 11 , wherein the entry is created in response to detecting a threshold number of interrupt signals from the I/O thread to the one of the virtual processors.
14. The system of claim 10 , wherein the processor is further to cause the system to store the map in a hypervisor memory.
15. The system of claim 10 , wherein the map associates the I/O thread with the virtual processor and a virtual machine corresponding to the virtual processor.
16. The system of claim 10 , wherein to cause the scheduler to move the I/O thread up in the processing queue, the processor is further to cause the system to lend execution time allocated to the one of the virtual processors to the I/O thread.
17. The system of claim 10 , wherein the processor is further to cause the system to:
detect interrupt signals from other I/O threads sent to other virtual processors of the number of virtual processors being executed on the computing system; and
map the other I/O threads to the other virtual processors.
18. A method performed by a physical computing system, the method comprising:
with a hypervisor, monitoring interrupt signals sent to virtual processors;
creating a map between Input/Output (I/O) threads associated with the interrupt signals and the virtual processors;
with a scheduler, managing a processing queue for a set of threads to be executed on a host processor, the set of threads including the I/O threads; and
in response to determining that a first virtual processor of the virtual processors is in an idle state;
using the map to find the I/O thread associated with the first virtual processor; and
causing the scheduler to use execution time allocated to the first virtual processor for execution of the I/O thread associated with the first virtual processor.
19. The method of claim 18 , wherein the first virtual processor is idle because the virtual processor executed a halt command.
20. The method of claim 18 , wherein creating the map comprises, for each detected interrupt signal, create an entry in a data structure that links the I/O thread that sent that interrupt signal to the virtual processor to which that interrupt signal is being sent.