Granted Patent B2

US 11,593,162 · App. 17/074,951 · Granted Feb 28, 2023

Inventors: Lior Kamran (Richon LeZion, IL); Amitai Alkalay (Kadima, IL); Liran Loya (Hod Hasharon, IL)

Assignee: EMC IP Holding Company LLC

G06F9/4881G06F9/485

Abstract

Claims (44)

1. A method of managing operation of a computing device, the method comprising:

running a system scheduler that schedules execution of a first application and a second application on a central processing unit (CPU) core of the computing device, the first application including an internal scheduler distinct from the system scheduler, executing the first application on the core including initiating, by the internal scheduler, a plurality of fibers of the first application on the core in series;

while the first application is executing on the core, detecting, by the internal scheduler of the first application in response to termination of execution of one of the plurality of fibers of the first application on the core, a context-switch opportunity; and

issuing, by the first application in response to detecting the context-switch opportunity, a blocking operation that triggers the system scheduler to perform a rescheduling operation between the first and second applications on the CPU core,

wherein the system scheduler is configured with a minimum granularity period (MGP), the MGP being a shortest length of time that the system scheduler is able to schedule, and wherein detecting the context-switch opportunity is done upon executing the first application on the core for less than the MGP,

wherein the one of the plurality of fibers is an initial fiber of the first application executing on the core that terminates at least a threshold amount of time after executing the first application on the core, the threshold amount of time being less than the MGP.

2. The method of claim 1 ,

wherein the MGP is within a range of 1 millisecond (ms) to 4 ms; and

wherein the threshold amount of time is within a range of 2.5% to 90% of the MGP.

3. The method of claim 1 ,

wherein the threshold amount of time is variable;

wherein the blocking operation is a sleep operation with a fixed sleep period; and

wherein the method further comprises:

after issuing the blocking operation, resuming execution of the first application on the core,

upon resuming execution, checking an elapsed time since issuing the blocking operation, and

in response to the elapsed time exceeding the fixed sleep period by more than an excess threshold, increasing the variable threshold amount of time.

4. The method of claim 3 ,

wherein the MGP is within a range of 1 millisecond (ms) to 4 ms; and

wherein the threshold amount of time has a default value within a range of 5% to 25% of the MGP.

5. The method of claim 1 ,

wherein the first application is a storage processing application configured to process block-level storage requests; and

wherein the plurality of fibers of the first application are polling fibers configured to poll storage resources managed by the storage processing application without issuing blocking operations.

6. The method of claim 5 wherein the second application is a storage processing application configured to process file-level storage requests.

7. The method of claim 5 wherein the second application does not operate in a polling mode, the second application being configured to issue blocking operations as part of reading and writing data.

8. The method of claim 1 wherein detecting the context-switch opportunity includes detecting that a container thread of the first application executing on the core does not hold a lock on any resource shared with another application running on the computing device or with another container thread of the first application executing on another core of the computing device.

9. The method of claim 1 wherein issuing the blocking operation includes issuing a sleep operation.

10. The method of claim 9 wherein the sleep operation has a period within a range of 250 nanoseconds to 5 microseconds.

11. The method of claim 1 wherein the context-switch opportunity is not triggered by a blocking operation of the first application.

12. A computer program product comprising a non-transitory computer-readable storage medium storing a set of instructions, which, when performed by a computing device, causes the computing device to:

run a system scheduler that schedules execution of a first application and a second application on a central processing unit (CPU) core of the computing device, the first application including an internal scheduler distinct from the system scheduler, execution of the first application on the core including initiation, by the internal scheduler, of a plurality of fibers of the first application on the core in series;

while the first application is executing on the core, detect, by the internal scheduler of the first application in response to termination of execution of one of the plurality of fibers of the first application on the core, a context-switch opportunity; and

issue, by the first application in response to detecting the context-switch opportunity, a blocking operation that triggers the system scheduler to perform a rescheduling operation between the first and second applications on the CPU core,

wherein the system scheduler is configured with a minimum granularity period (MGP), the MGP being a shortest length of time that the system scheduler is able to schedule, and wherein detecting the context-switch opportunity is done upon executing the first application on the core for less than the MGP,

wherein the one of the plurality of fibers is an initial fiber of the first application configured to execute on the core that terminates at least a threshold amount of time after execution of the first application on the core, the threshold amount of time being less than the MGP.

13. The computer program product of claim 12 wherein detection of the context-switch opportunity includes detection that a container thread of the first application configured to execute on the core does not hold a lock on any resource shared with another application configured to run, on the computing device or with another container thread of the first application configured to execute on another core of the computing device.

14. The computer program product of claim 12 wherein issuance of the blocking operation includes issuance of a sleep operation.

15. An apparatus comprising:

a central processing unit (CPU) including at least one core; and

memory, the memory storing instructions, which, when performed by the core cause the core to:

run a system scheduler that schedules execution of a first application and a second application on the core, the first application including an internal scheduler distinct from the system scheduler, execution of the first application on the core including initiation, by the internal scheduler, of a plurality of fibers of the first application on the core in series;

while the first application is executing on the core, detect, by the internal scheduler of the first application in response to termination of execution of one of the plurality of fibers of the first application on the core, a context-switch opportunity; and

issue, by the first application in response to detecting the context-switch opportunity, a blocking operation that triggers the system scheduler to perform a rescheduling operation between the first and second applications on the core,

wherein the system scheduler is configured with a minimum granularity period (MGP), the MGP being a shortest length of time that the system scheduler is able to schedule, and wherein detection of the context-switch opportunity is done upon executing the first application on the core for less than the MGP,

wherein the one of the plurality of fibers is an initial fiber of the first application configured to execute on the core that terminates at least a variable threshold amount of time after execution of the first application on the core, the variable threshold amount of time being less than the MGP.

Assignments (9)

Continuity (1)