Temporal batching of I/O jobs

• US 8,245,229 B2
• Filed: 09/30/2008
• Issued: 08/14/2012
• Est. Priority Date: 09/30/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Green, Dustin L., Chin, Yau Ning, Worthington, Bruce L., Ganguly, Shuvabrata, Garg, Pankaj
• Primary Examiner(s): Coleman, Eric
• Application Number:
  US12/242,615
• Publication Number:
  US 20100083256A1
• Time in Patent Office: 1,414 Days
• Field of Search: None
• US Class Current: 718/101
• CPC Class Codes: G06F 2209/5019 Workload prediction
• CPC Class Codes: G06F 2209/5022 Workload threshold
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system

Abstract

Batching techniques are provided to maximize the throughput of a hardware device based on the saturation point of the hardware device. A balancer can determine the saturation point of the hardware device and determine the estimated time cost for IO jobs pending in the hardware device. A comparison can be made and if the estimated time cost total is lower than the saturation point one or more IO jobs can be sent to the hardware device.

Claims

1. An IO job batching method, comprising:

determining a saturation point of a hardware device;

estimating a time cost for IO jobs pending in the hardware device and determining that the estimated time cost total to a time cost value that is less than the saturation point; and

sending an amount of IO jobs to the hardware device, the amount of IO jobs increasing the estimated time cost for pending IO jobs past at least the saturation point.

2. The method of claim 1, wherein the amount of IO jobs sent to the hardware device increases the estimated time cost for IO jobs pending in the hardware device past an upper saturation point.

3. The method of claim 1, further comprising:

determining that the estimated time cost for IO jobs pending in the hardware device is greater than the saturation point; and

delaying IO jobs.

4. The method of claim 1, further comprising:

determining that the estimated time cost for IO jobs pending in the hardware is lower than a lower saturation point; and

sending subsequent IO jobs to the hardware device upon receipt of the subsequent IO jobs.

5. The method of claim 1, wherein the saturation point is associated with maximum throughput of the hardware device.

6. The method of claim 1, wherein the saturation point is set based on an estimated time cost of an average IO job.

7. The method of claim 1, wherein the saturation point is dynamically determined based on operational characteristics of the hardware device.

8. The method of claim 1, further comprising:

accessing a memory location, the memory location shared with a virtual machine;

retrieving an IO job from the shared memory location, the IO job previously stored in the memory location by the virtual machine; and

disabling inter-partition notifications associated with the shared memory location, wherein inter-partition notifications are used by the virtual machine to indicate that IO jobs are stored in the shared memory location.

9. A computer readable storage medium including computer executable instructions for batching IO jobs, the computer readable storage medium comprising:

instructions for receiving IO jobs associated with a client;

instructions for allocating a quantum amount of time that a hardware device can spend processing client IO jobs, the quantum amount of time based on an estimated amount of time required by the hardware device to perform a seek operation; and

instructions for sending client IO jobs to the hardware device, the amount of client IO jobs having an estimated time cost equal to or less than the allocated quantum amount of time.

10. The computer readable storage medium of claim 9, further comprising:

instructions for reducing the allocated quantum amount of time in proportion to an estimated time cost for the client IO jobs sent to the hardware.

11. The computer readable storage medium of claim 9, further comprising:

instructions for determining that estimated time cost for IO jobs pending in the hardware device totals to a time cost value that is less than a saturation point.

12. The computer readable storage medium of claim 9, further comprising:

instructions for accessing a memory location, the memory location shared with a virtual machine;

instructions for retrieving an IO job from the shared memory location, the IO job previously stored in the memory location by the virtual machine; and

instructions for disabling inter-partition notifications associated with the shared memory location, wherein inter-partition notifications are used by the virtual machine to indicate that IO jobs are stored in the shared memory location.

13. The computer readable storage medium of claim 9, further comprising:

instructions for associating the client with a total amount of time that the hardware device can spend processing client IO jobs.

14. The computer readable storage medium of claim 13, further comprising:

instructions for allocating the quantum amount of time that the hardware device can spend processing client IO jobs from the total amount of time.

15. The computer readable storage medium of claim 14, further comprising:

instructions for reducing the total amount of time that the hardware device can spend processing client IO jobs; and

instructions for storing subsequent client IO jobs in a queue.

16. A computer system configured to batch IO jobs, the computer system comprising:

circuitry for storing a time value associated with a client, the time value being in a delay state, the delay state indicating that the hardware device was overused by the client;

circuitry for receiving a job completion signal from a hardware device;

circuitry for increasing the time value an amount of time to increase the value to an issue state and an additional amount that is based on a saturation point for the hardware device; and

circuitry for sending IO jobs to the hardware device that are eligible based on the increased time value.

17. The computer system of claim 16, wherein the saturation point is associated with maximum throughput of the hardware device.

18. The computer system of claim 16, wherein sending IO jobs to the hardware device that are eligible further comprises:

circuitry for allocating an amount of time that a hardware device can spend processing client IO jobs, the amount of time based on an estimated amount of time required by the hardware device to perform a seek operation; and

circuitry for reducing the time value in proportion to estimated time cost values for the IO jobs that were sent to the hardware device.

19. The computer system of claim 16, further comprising:

circuitry for determining that estimated time cost values for IO jobs pending in the hardware device totals to a time cost value that is less than a saturation point.

20. The computer system of claim 16, further comprising:

circuitry for accessing a memory location, the memory location shared with a virtual machine;

circuitry for retrieving an IO job from the shared memory location, the IO job previously stored in the memory location by the virtual machine; and

circuitry for disabling inter-partition notifications associated with the shared memory location, wherein inter-partition notifications are used by the virtual machine to indicate that IO jobs are stored in the shared memory location.