Method and system for heuristic throttling for distributed file systems

• US 8,275,902 B2
• Filed: 09/22/2008
• Issued: 09/25/2012
• Est. Priority Date: 09/22/2008
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Oracle America, Inc. (Oracle Corporation)
• Inventors: Philip, Binu J., Salam, Sudheer Abdul
• Primary Examiner(s): Najjar, Saleh
• Assistant Examiner(s): Shepperd, Eric W
• Application Number:
  US12/235,106
• Publication Number:
  US 20100077096A1
• Time in Patent Office: 1,464 Days
• Field of Search: 709/203, 709/217, 709219-220, 709223-224, 709/226, 709/230, 709232-235, 710/29, 710/52, 710/58, 710/60, 370229-234
• US Class Current: 709/233
• CPC Class Codes: G06F 16/182 Distributed file systems

Abstract

A method for throttling in a file system. The method includes receiving a request by a client from an application to write data to a server, wherein the application is associated with an application bandwidth and wherein the server is associated with a server bandwidth, queuing a first job to write the data to the server on an I/O queue using an application-to-client throttling mechanism using at least one selected from the group of the application bandwidth and the server bandwidth, and issuing the first job from the I/O queue to the server using a client-to-server throttling mechanism and at least one selected from the group of the application bandwidth and the server bandwidth.

Claims

1. A method for throttling in a file system, comprising:

receiving a first request by a client from an application to write first data to a server across a network, wherein the application is associated with an application bandwidth comprising a rate at which the application writes data to a client storage device;

storing the first data in the client storage device using the application bandwidth;

queuing a first I/O request to write the first data to the server on an I/O queue;

issuing the first I/O request from the I/O queue to the server using a server bandwidth comprising a rate at which I/O requests to write data to the server are issued;

receiving an acknowledgment that the first I/O request has been completed by the server;

determining a first server acknowledgement rate using the acknowledgment that the first I/O request has been completed by the server;

determining a first perceived bandwidth using the application bandwidth, the first server acknowledgement rate, and an amount of available space on the client storage device;

resetting the server bandwidth to the perceived bandwidth to obtain a reset server bandwidth;

receiving a second request by the client from the application to write second data to the server;

storing the second data in the client storage device using the application bandwidth;

queuing a second I/O request to write the second data to the server on the I/O queue; and

issuing the second I/O request from the I/O queue to the server using the reset server bandwidth.

2. The method of claim 1, further comprising:

receiving an acknowledgment that the second I/O request has been completed by the server;

determining a second server acknowledgement rate using the acknowledgment that the second I/O request has been completed by the server;

determining a second perceived bandwidth using the application bandwidth, the second server acknowledgement rate, and the amount of available space on the client storage device; and

resetting the application bandwidth to the second perceived bandwidth.

3. A non-transitory computer readable storage medium for throttling in a file system, comprising computer readable code, wherein the code is configured to:

receive a request, by a client from an application, to write first data to a server across a network, wherein the application is associated with an application bandwidth comprising a rate at which the application writes data to a client storage device;

store the first data in the client storage device using the application bandwidth;

queue a first I/O request to write the first data to the server on an I/O queue;

issue the first I/O request from the I/O queue to the server using a server bandwidth comprising a rate at which I/O requests to write data to the server are issued;

receive an acknowledgment that the first I/O request has been completed by the server;

determine a first server acknowledgement rate using the acknowledgment that the first I/O request has been completed by the server,

determine a first perceived bandwidth using the application bandwidth, the first server acknowledgement rate, and an amount of available space on the client storage device;

reset the server bandwidth to the perceived bandwidth to obtain a reset server bandwidth;

receive a second request by the client from the application to write second data to the server;

store the second data in the client storage device using the application bandwidth;

queue a second I/O request to write the second data to the server on the I/O queue; and

issue the second I/O request from the I/O queue to the server using the reset server bandwidth.

4. The non-transitory computer readable storage medium of claim 3, wherein the code is further configured to:

receive an acknowledgment that the second I/O request has been completed by the server;

determine a second server acknowledgement rate using the acknowledgment that the second I/O request has been completed by the server;

determine a second perceived bandwidth using the application bandwidth, the second server acknowledgement rate, and the amount of available space on the client storage device; and

reset the application bandwidth to the second perceived bandwidth to obtain a reset application bandwidth.

5. The non-transitory computer readable storage medium of claim 4, wherein the code is further configured to, upon resetting the application bandwidth:

receiving a request by the client from the application to write data to the server;

storing the third data in the client storage device;

queuing a third I/O request to write the third data to the server on the I/O queue using the reset application bandwidth; and

issuing the third I/O request from the I/O queue to the server using the reset server bandwidth.

6. The non-transitory computer readable storage medium of claim 3, wherein the code is further configured to:

make a first determination that a size of the first data is greater than the application bandwidth;

make a second determination that the amount of available space on the client storage device is sufficient;

make a third determination that the application bandwidth is greater than zero;

allocate space in the client storage device to store the first data based on the first determination, the second determination, and the third determination; and

write the first data to the allocated space in the client storage device.

7. The non-transitory computer readable storage medium of claim 3, wherein the code is further configured to:

determine that a size of the first data is greater than the application bandwidth;

determine that the amount of available space on the client storage device is insufficient; and

write the first data to the client storage device when the application bandwidth is available.

8. The non-transitory computer readable storage medium of claim 3, wherein the code is further configured to:

determine that a size of the first data is not greater than the application bandwidth;

allocate space in the client storage device to store the first data; and

write the first data to the allocated space in the client storage device when the application bandwidth is available.

9. The non-transitory computer readable storage medium of claim 3, wherein the code is further configured to:

determine that a size of the first data is not greater than the server bandwidth; and

decrement the server bandwidth.

10. The non-transitory computer readable storage medium of claim 3, wherein the code is further configured to:

make a first determination that a size of the first data is greater than the server bandwidth;

make a second determination that the amount of available space on the client storage device is insufficient;

modify the first I/O request to write the first data synchronously based on the first determination and the second determination; and

decrement the server bandwidth.

11. The non-transitory computer readable storage medium of claim 3, wherein the client is part of a client cluster.

12. A client, comprising:

a processor;

an application, executing on the processor, configured to send a request to the client to write first data to a server, wherein the application is associated with an application bandwidth and wherein the server is associated with a server bandwidth;

a client storage device configured to store the first data using the application bandwidth comprising a rate at which the application writes data to a client storage device;

an I/O queue configured to;

queue a first I/O request to write the first data to the server; and

issue the first I/O request to the server using a server bandwidth comprising a rate at which I/O requests to write data to the server are issued;

an acknowledgement engine configured to receive an acknowledgment that the first I/O request has been completed by the server;

a bandwidth engine configured to;

determine a first server acknowledgement rate using the acknowledgment that the first I/O request has been completed by the servers;

determine a first perceived bandwidth using the application bandwidth, the first server acknowledgement rate, and an amount of available space on the client storage device; and

reset the server bandwidth to the perceived bandwidth to obtain a reset server bandwidth;

wherein the application is further configured to send a second request to the client to write second data to the server;

wherein the client storage device is further configured to store the second data using the application bandwidth; and

wherein the I/O queue is further configured to;

queue a second I/O request to write the second data to the server; and

issue the second I/O request to the server using the reset server bandwidth.

13. The client of claim 12, wherein the bandwidth engine is further configured to:

receive an acknowledgment that the second I/O request has been completed by the server;

determine a second server acknowledgement rate using the acknowledgment that the second I/O request has been completed by the server;

determine a second perceived bandwidth using the application bandwidth, the second server acknowledgement rate, and the amount of available space on the client storage device; and

resetting the application bandwidth to the second perceived bandwidth to obtain a reset application bandwidth.

14. The client of claim 13,

wherein the bandwidth engine is further configured to, upon resetting the application bandwidth:

receive a request from the application to write third data to the server;

store the third data in the client storage device; and

queue a third I/O request to write the third data to the server on the I/O queue using the reset application bandwidth; and

wherein the I/O queue is further configured to issue the third I/O request to the server using the reset server bandwidth.

15. The client of claim 12, wherein the bandwidth engine comprises an application-to-client throttling mechanism configured to:

determine that the amount of available space on the client storage device is sufficient;

determine that the application bandwidth is greater than zero;

allocate space in the client storage device to store the first data; and

write the first data to the allocated space in the client storage device.

16. The client of claim 12, wherein the bandwidth engine comprises an application-to-client throttling mechanism configured to:

determine that a size of the first data is greater than the application bandwidth;

determine that the amount of available space on the client storage device is insufficient; and

write the first data to the client storage device when the application bandwidth is available.

17. The client of claim 12, wherein the bandwidth engine comprises an application-to-client throttling mechanism configured to:

determine that a size of the first data is not greater than the application bandwidth;

allocate space in the client storage device to store the first data; and

write the first data to the allocated space in the client storage device when the application bandwidth is available.

18. The client of claim 12, wherein the bandwidth engine comprises a client-to-server throttling mechanism configured to:

determine that a size of the first data is not greater than the server bandwidth; and

decrement the server bandwidth.

19. The client of claim 12, wherein the bandwidth engine comprises a client-to-server throttling mechanism configured to:

determine that a size of the first data is greater than the server bandwidth;

determine that the amount of available space on the client storage device is insufficient;

modify the first I/O request to write the first data synchronously; and

decrement the server bandwidth.

20. The client of claim 12, wherein the client is part of a client cluster.