SPECULATIVE AND COORDINATED DATA ACCESS IN A HYBRID MEMORY SERVER

• US 20110320523A1
• Filed: 06/24/2010
• Published: 12/29/2011
• Est. Priority Date: 06/24/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Krishnamurthy, Rajaram B., Parris, Carl Joseph, CHAN, Yuk Lung
• Granted Patent: US 8,954,490 B2
• Time in Patent Office: Days
• US Class Current: 709/203
• CPC Class Codes: G06F 16/14 Details of searching files based on file metadata
• CPC Class Codes: G06F 16/172 Caching, prefetching or hoarding of files
• CPC Class Codes: G06F 16/182 Distributed file systems

Abstract

A method, accelerator system, and computer program product, for prefetching data from a server system in an out-of-order processing environment. A plurality of prefetch requests associated with one or more given data sets residing on the server system are received from an application on the server system. Each prefetch request is stored in a prefetch request queue. A score is assigned to each prefetch request. A set of the prefetch requests are selected from the prefetch queue that comprise a score above a given threshold. A set of data, for each prefetch request in the set of prefetch requests, is prefetched from the server system that satisfies each prefetch request, respectively.

Claims

1. A method, on a server system in an out-of-core processing environment, for speculatively managing data at the server system, the method comprising:

receiving a request to access a given data set from a user client;

identify, based on the request, at least a portion of the given data set that satisfies the request;

retrieving at least the portion of the given data set that has been identified; and

passing at least one of the request and at least the portion of the given data set that has been identified to one of an accelerator system and the user client.

2. The method of claim 1, wherein the receiving further comprises:

establishing a tunnel protocol between the user client and the accelerator system.

3. The method of claim 1, further comprising:

receiving from the accelerator system a pull request for data associated with the given data set.

4. The method of claim 3, further comprising:

identifying at least a portion of the given data set that satisfies the pull request;

determining that at least the portion of the given data is associated with a set of related data; and

sending at least the portion of the given data and the set of related data to the accelerator system.

5. The method of claim 4, wherein determining that at least the portion of the given data is associated with a set of related data further comprises:

analyzing a set of semantic information associated with a current state of an application utilizing the given data set; and

selecting, in response to the analyzing, the set of related data from a larger set of related data.

6. The method of claim 1, further comprising:

determining that the user client is able to process the portion of the given data set; and

sending the portion of the given data set directly to the user client.

7. The method of claim 1, further comprising:

determining that an amount of data to satisfy the request to access is above a given threshold; and

passing the request along with at least the portion of the given data set that has been identified to the accelerator system.

8. The method of claim 1, further comprising:

determining that an amount of data to satisfy the request to access is below a given threshold; and

passing at least the portion of the given data set that has been identified to the user client.

9. A method, with an accelerator system in an out-of-order processing environment, for prefetching data from a server system in the out-of-order processing environment, the method comprising:

receiving, from an application on at least one of the server system and a user client, a plurality of prefetch requests associated with one or more given data sets residing on the server system;

storing each prefetch request in a prefetch request queue;

assigning a score to each prefetch request;

selecting a set of the prefetch requests from the prefetch queue that comprise a score above a given threshold; and

prefetching, for each prefetch request in the set of prefetch requests, a set of data from the server system that satisfies each prefetch request, respectively.

10. The method of claim 9, further comprising:

storing the set of data in a portion of memory that is separate from non-prefetched data.

11. The method of claim 9, wherein assigning a score is based at least on determining a set of resources required by the prefetch request.

12. The method of claim 9, further comprising:

determining, in response to selecting the set of prefetch requests, that at least two prefetch requests in the set of prefetch requests are requesting substantially similar data; and

aggregating the at least two prefetch requests into a single prefetch request.

13. The method of claim 9, wherein selecting the set of prefetch requests further comprises:

identifying a current state of the application; and

selecting at least one prefetch request based on the current state of the application that has been identified.

14. A server system in an out-of-core processing environment for speculatively managing data at the server system, the server system comprising:

a memory;

a processor communicatively coupled to the memory; and

a data access manager communicatively coupled to the memory and the processor, the data access manager configured to perform a method comprising;

receiving a request to access a given data set from a user client;

identify, based on the request, at least a portion of the given data set that satisfies the request;

retrieving at least the portion of the given data set that has been identified; and

passing at least one of the request and at least the portion of the given data set that has been identified to one or the an accelerator system and the user client.

15. The server system of claim 14, wherein the receiving further comprises:

establishing a tunnel protocol between the user client and the accelerator system.

16. The server system of claim 14, wherein the method further comprises:

receiving from the accelerator system a pull request for data associated with the given data set.

17. The server system of claim 16, wherein the method further comprises:

identifying at least a portion of the given data set that satisfies the pull request;

determining that at least the portion of the given data is associated with a set of related data; and

sending at least the portion of the given data and the set of related data to the accelerator system.

18. The server system of claim 17, wherein determining that at least the portion of the given data is associated with a set of related data further comprises:

analyzing a set of semantic information associated with a current state of an application utilizing the given data set; and

selecting, in response to the analyzing, the set of related data from a larger set of related data.

19. An accelerator system in an out-of-order processing environment for prefetching data from a server system in the out-of-order processing environment, the accelerator system comprising:

a memory;

a processor communicatively coupled to the memory; and

a data access manager communicatively coupled to the memory and the processor, the data access manager configured to perform a method comprising;

receiving, from an application on at least one of the server system and a user client, a plurality of prefetch requests associated with one or more given data sets residing on the server system;

storing each prefetch request in a prefetch request queue;

assigning a score to each prefetch request;

selecting a set of the prefetch requests from the prefetch queue that comprise a score above a given threshold; and

prefetching, for each prefetch request in the set of prefetch requests, a set of data from the server system that satisfies each prefetch request, respectively.

20. The accelerator system of claim 19, wherein assigning a score is based at least on determining a set of resources required by the prefetch request.

21. The accelerator system of claim 19, wherein the method further comprises:

determining, in response to selecting the set of prefetch requests, that at least two prefetch requests in the set of prefetch requests are requesting substantially similar data; and

aggregating the at least two prefetch requests into a single prefetch request.

22. A computer program product for prefetching data from a server system in the out-of-order processing environment, the computer program product comprising:

a storage medium readable by a processing circuit and storing instructions for execution by the processing circuit for performing a method comprising;

receiving, from an application on the server system, a plurality of prefetch requests associated with one or more given data sets residing on the server system;

storing each prefetch request in a prefetch request queue;

assigning a score to each prefetch request;

selecting a set of the prefetch requests from the prefetch queue that comprise a score above a given threshold; and

prefetching, for each prefetch request in the set of prefetch requests, a set of data from the server system that satisfies each prefetch request, respectively.

23. The computer program product of claim 22, wherein assigning a score is based at least on determining a set of resources required by the prefetch request.

24. The computer program product of claim 22, wherein the method further comprises:

determining, in response to selecting the set of prefetch requests, that at least two prefetch requests in the set of prefetch requests are requesting substantially similar data; and

aggregating the at least two prefetch requests into a single prefetch request.

25. The computer program product of claim 22, wherein selecting the set of prefetch requests further comprises:

identifying a current state of the application; and

selecting at least one prefetch request based on the current state of the application that has been identified.