System and method for supporting adaptive busy wait in a computing environment

• US 9,558,035 B2
• Filed: 08/01/2014
• Issued: 01/31/2017
• Est. Priority Date: 12/18/2013
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Falco, Mark A.
• Primary Examiner(s): Puente, Emerson
• Assistant Examiner(s): Kamran, Mehran
• Application Number:
  US14/449,357
• Publication Number:
  US 20150169367A1
• Time in Patent Office: 914 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 1/3296 by lowering the supply or operating voltage
• CPC Class Codes: G06F 9/485 Task life-cycle, e.g. stopping, restarting, resuming execution G06F9/4881 takes precedence
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes
• CPC Class Codes: G06F 9/546 Message passing systems or structures, e.g. queues

Abstract

A system and method can support queue processing in a computing environment such as a distributed data grid. A thread can be associated with a queue in the computing environment, wherein the thread runs on one or more microprocessors that support a central processing unit (CPU). The system can use the thread to process one or more tasks when said one or more tasks arrive at the queue. Furthermore, the system can configure the thread to be in one of a sleep state and an idle state adaptively, when there is no task in the queue.

Claims

1. A method for supporting queue processing in a computing environment, the method comprising:

associating a single thread with a queue in the computing environment, wherein the thread runs on one or more microprocessors;

using the thread to process a next task from said queue if said queue is not empty;

waiting for a next task to be received by said queue if said queue is empty;

adaptively configuring the thread, responsive to a workload of said queue, to be in one of a sleep state and an idle state during said waiting;

associating a single another thread with another queue in the computing environment, wherein the another thread runs on the one or more microprocessors;

using the another thread to process another next task from said another queue if said another queue is not empty;

waiting for another next task to be received by said another queue if said another queue is empty; and

adaptively configuring the another thread, responsive to another workload of said another queue, to be in one of a sleep state and an idle state during said waiting;

wherein each of said thread and said another thread are adaptively configured independently to be in the respective sleep state or idle state during said respective waiting, responsive to the workload of said queue and the another workload of said another queue.

2. The method according to claim 1, wherein adaptively configuring the thread, responsive to a workload of said queue, to be in one of a sleep state and an idle state, during said waiting comprises:

using an operating system function to determine a total sleep state cost for putting the thread in the sleep state and awakening the thread from the sleep state;

configuring the thread to be in said sleep state during said waiting if said total sleep state cost is less than an idle state cost for configuring the thread to be in said idle state during said waiting; and

configuring the thread to be in said idle state during said waiting if said total sleep state cost is not less than said idle state cost for configuring the thread to be in said idle state during said waiting.

3. The method according to claim 1, further comprising:

awakening the thread from the sleep state when a new task arrives at the queue if said thread is in said sleep state during said waiting period, and the processing said new task with said thread; and

processing said new task immediately with said thread if said thread is in said idle state during said waiting period.

4. The method according to claim 1, wherein adaptively configuring the thread, responsive to a workload of said queue, to be in one of a sleep state and an idle state, during said waiting comprises:

obtaining a total sleep state cost for putting the thread in the sleep state and awakening the thread from the sleep state, based on an operating system (OS) function;

determining based on said workload of said queue an idle state cost for configuring the thread to be in said idle state during said waiting;

comparing said total sleep state cost with said idle state cost;

configuring the thread to be in said sleep state during said waiting if said total sleep state cost is less than said idle state cost; and

configuring the thread to be in said idle state during said waiting if said total sleep state cost is not less than said idle state cost.

5. The method according to claim 1, wherein said idle state comprises a busy wait state in which the thread utilizes computing resources of said one or more microprocessor during said waiting.

6. The method according to claim 1, wherein adaptively configuring the thread, responsive to a workload of said queue, to be in one of a sleep state and an idle state, during said waiting comprises:

obtaining a total sleep state cost for putting the thread in the sleep state and awakening the thread from the sleep state, based on an operating system (OS) function;

determining an idle state cost for configuring the thread to be in said idle state during said waiting based on an average interval of time between consecutive tasks arriving at said queue;

comparing said total sleep state cost with said idle state cost;

configuring the thread to be in said sleep state during said waiting if said total sleep state cost is less than said idle state cost; and

configuring the thread to be in said idle state during said waiting if said total sleep state cost is not less than said idle state cost.

7. The method according to claim 1, further comprising:

comparing a total sleep state cost for putting the thread in the sleep state and waking the thread up from the sleep state with an idle cost for putting the thread in the idle state wherein said idle state cost is based on an average interval of time between consecutive tasks arriving at said queue.

8. The method according to claim 7, further comprising:

configuring the thread to be in a sleep state if the total sleep state cost for putting the thread in the sleep state is smaller than the idle state cost for putting the thread in the idle state; and

configuring the thread to be in an idle state if otherwise.

9. The method according to claim 6, wherein adaptively configuring the another thread, responsive to said another workload of said another queue, to be in one of a sleep state and an idle state, during said waiting comprises:

obtaining another total sleep state cost for putting the another thread in the sleep state and awakening the another thread from the sleep state, based on the operating system (OS) function;

determining another idle state cost for configuring the another thread to be in said idle state during said waiting based on an average interval of time between consecutive tasks arriving at said another queue;

comparing said another total sleep state cost with said another idle state cost;

configuring the another thread to be in said sleep state during said waiting if said another total sleep state cost is less than the another idle state cost; and

configuring the another thread to be in said idle state during said waiting if said another total sleep state cost is not less than said another idle state cost.

10. The method according to claim 8, further comprising:

comparing another total sleep state cost for putting the another thread in the sleep state and waking the another thread up from the sleep state with another idle cost for putting the another thread in the idle state wherein said idle state cost is based on an average interval of time between consecutive tasks arriving at said another queue;

configuring the another thread to be in a sleep state if the another total sleep state cost for putting the another thread in the sleep state is smaller than the another idle state cost for putting the another thread in the idle state; and

configuring the another thread to be in an idle state if otherwise.

11. A system for supporting task processing in a distributed data grid, the system comprising:

a computer system comprising one or more microprocessors and memory;

a queue in said memory and another queue in said memory;

a thread, running on the one or more microprocessors, wherein the thread is associated with said queue, wherein the thread is configured to

process a next task from said queue if said queue is not empty;

wait for a next task to be received by said queue if said queue is empty; and

enter one of a sleep state and an idle state while waiting for the next task, wherein selecting between the sleep state and the idle state is performed adaptively responsive to a workload of said queue;

another thread, running on the one or more microprocessors, wherein the another thread is associated with said another queue, wherein the another thread is configured to

process another next task from said another queue if said another queue is not empty;

wait for another next task to be received by said another queue if said another queue is empty; and

enter one of a sleep state and an idle state while waiting for the another next task, wherein selecting between the sleep state and the idle state is performed adaptively responsive to the another workload of said another queue;

wherein each of said thread and said another thread are configured to independently enter the respective sleep state or idle state during said respective waiting, responsive to the workload of said queue and the another workload of said another queue.

12. The system according to claim 11, wherein:

said thread is configured to enter the sleep state while waiting for a next task when the workload of the queue is light and enter the idle state while waiting for a next task when the workload of the queue is heavy.

13. The system according to claim 11, wherein:

said thread is configured to enter the sleep state while waiting for a next task when an average time between consecutive tasks received at said queue is long and enter the idle state while waiting for a next task when task when an average time between consecutive tasks received at said queue is short.

14. The system according to claim 11, wherein:

the thread is configured to obtain a total cost for putting the thread in the sleep state and waking the thread up from the sleep state using an operating system (OS) function.

15. The system according to claim 11, wherein:

said idle state comprises a busy wait state in which the thread utilizes computing resources of said one or more microprocessor while waiting for a next task.

16. The system according to claim 11, wherein the thread is configured to select between the idle state and the slap state by:

obtaining a total sleep state cost for putting the thread in the sleep state and awakening the thread from the sleep state, based on an operating system (OS) function;

determining an idle state cost for configuring the thread to be in said idle state during said waiting based on an average interval of time between consecutive tasks arriving at said queue;

comparing said total sleep state cost with said idle state cost;

enter said sleep state while waiting for the next task if said total sleep state cost is less than said idle state cost; and

enter said idle state while waiting for said next task if said total sleep state cost is not less than said idle state cost.

17. The system according to claim 11, wherein the thread is configured to select between the idle state and the sleep state by:

comparing a total sleep state cost for putting the thread in the sleep state and awakening the thread from the sleep state with an idle state cost for putting the thread in the idle state.

18. The system according to claim 17, wherein:

the thread is configured to enter the sleep state while waiting for the next task when the total sleep state cost for putting the thread in the sleep state is smaller than the idle state cost for putting the thread in the idle state; and

the thread is configured to enter the idle state while waiting for the next task if otherwise.

19. The system according to claim 12, wherein:

said another thread is configured to enter the sleep state while waiting for the another next task when the another workload of the another queue is light and enter the idle state while waiting for the another next task when the another workload of the another queue is heavy.

20. A non-transitory machine readable storage medium having instructions stored thereon for supporting queue processing in a computing environment, which instructions, when executed, cause a computer to perform steps comprising:

associating a single thread with a queue in the computing environment, wherein the thread runs on one or more microprocessors;

using the thread to process a next task from said queue if said queue is not empty;

waiting for a next task to be received by said queue if said queue is empty;

adaptively configuring the thread, responsive to a workload of said queue, to be in one of a sleep state and an idle state during said waiting;

associating a single another thread with another queue in the computing environment, wherein the another thread runs on the one or more microprocessors;

using the another thread to process another next task from said another queue if said another queue is not empty;

waiting for another next task to be received by said another queue if said another queue is empty; and

adaptively configuring the another thread, responsive to another workload of said another queue, to be in one of a sleep state and an idle state during said waiting;

wherein each of said thread and said another thread are adaptively configured independently to be in the respective sleep state or idle state during said respective waiting, responsive to the workload of said queue and the another workload of said another queue.