Method and system for storage-aware flow resource management

• US 20040111527A1
• Filed: 12/06/2002
• Published: 06/10/2004
• Est. Priority Date: 12/06/2002
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Voruganti, Kaladhar, Uttamchandani, Sandeep Madhav, Sarkar, Prasenjit, Czap, Joseph C. Jr.
• Granted Patent: US 7,421,502 B2
• Time in Patent Office: Days
• US Class Current: 709/235
• CPC Class Codes: G06F 9/5011 the resources being hardware resources other than CPUs, Servers and Terminals
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system

Abstract

A method and structure for managing resources in a storage system that is based on server resource usage is disclosed. The method allocates a command window to each of the servers accessing the resources, achieves an equilibrium state that balances the size of each server'"'"'s command window and the amount of free resources, and maintains the equilibrium state by continually adjusting the size of each command window. The command window defines a limit on the number of outstanding storage requests that each server may maintain. The invention initially allocates a minimum sized command window.

Claims

1. A method for managing resources in a storage system being accessed by servers based on server resource usage, said method comprising:

allocating a command window to each of said servers accessing said resources;

achieving an equilibrium state that balances the size of each server'"'"'s command window and the amount of free resources; and

maintaining said equilibrium state by continually adjusting the size of each command window.

2. The method in claim 1, wherein said command window defines a limit on the number of outstanding storage requests that each server may maintain.

3. The method in claim 1, wherein said allocating initially allocates a minimum sized command window.

4. The method in claim 1, wherein said achieving of said equilibrium state comprises operating in an equilibrium achievement mode,

wherein, while operating in said equilibrium achievement mode, said method comprises increasing the size of each server'"'"'s command window at a first rate as each of said servers completes a successful storage operation.

5. The method in claim 4, wherein said first rate comprises an exponential rate.

6. The method in claim 4, wherein said maintaining of said equilibrium state comprises changing to a equilibrium maintenance mode when a storage request is unsuccessful,

wherein, while operating in said equilibrium maintenance mode, said method comprises:

decreasing the size of each command window by a reduction factor; and

increasing the size of each server'"'"'s command window at a second rate less than said first rate as each of said servers completes a successful storage operation.

7. The method in claim 6, wherein said second rate comprises a rate inversely proportional to the size of each command window.

8. The method in claim 6, further comprising:

reverting to said equilibrium achievement mode when the level of free resources rises above an upper limit; and

repeating said changing and reverting processes.

9. The method in claim 8, further comprising:

stopping said increasing processes when the amount of free resources falls below a lower limit;

restarting said increasing processes when the amount of said free resources rises above said lower limit; and

repeating said stopping and said restarting processes.

10. The method in claim 8, further comprising:

stopping said increasing processes for a server when said server stops utilizing said resources;

restarting said increasing processes when said server again begins utilizing said resources; and

repeating said stopping and said restarting processes.

11. A method for managing resources in a storage system being accessed by servers, said method comprising:

allocating a command window to each of said servers accessing said resources;

initially operating in an equilibrium achievement mode, wherein, while operating in said equilibrium achievement mode, said method comprises increasing the size of each server'"'"'s command window at a first rate as each of said servers completes a successful storage operation;

changing to a equilibrium maintenance mode when a storage request is unsuccessful, wherein, while operating in said equilibrium maintenance mode, said method comprises;

decreasing the size of each command window by a reduction factor; and

increasing the size of each server'"'"'s command window at a second rate less than said first rate as each of said servers completes a successful storage operation;

reverting to said equilibrium achievement mode when the level of free resources rises above an upper limit; and

repeating said changing and reverting processes.

12. The method in claim 11, wherein said repeating process acquires and maintains an equilibrium balance between the size of each command window and the amount of free resources.

13. The method in claim 11, wherein said command window defines a limit on the number of outstanding storage requests that each server may maintain.

14. The method in claim 11, wherein said allocating initially allocates a minimum sized command window.

15. The method in claim 11, further comprising:

stopping said increasing processes when the amount of free resources falls below a lower limit;

restarting said increasing processes when the amount of said free resources rises above said lower limit; and

repeating said stopping and said restarting processes.

16. The method in claim 11, further comprising:

stopping said increasing processes for a server when said server stops utilizing said resources;

restarting said increasing processes when said server again begins utilizing said resources; and

repeating said stopping and said restarting processes.

17. A method for managing resources in a storage system being accessed by servers, said method comprising:

allocating a command window to each of said servers accessing said resources,

initially operating in an equilibrium achievement mode, wherein, while operating in said equilibrium achievement mode, said method comprises increasing the size of each server'"'"'s command window at an exponential rate as each of said servers completes a successful storage operation;

changing to a equilibrium maintenance mode when a storage request is unsuccessful, wherein, while operating in said equilibrium maintenance mode, said method comprises;

decreasing the size of each command window by a reduction factor;

increasing the size of each server'"'"'s command window at a rate inversely proportional to the size of each command window as each of said servers completes a successful storage operation;

reverting to said equilibrium achievement mode when the level of free resources rises above an upper limit; and

repeating said changing and reverting processes.

18. The method in claim 17, wherein said repeating process acquires and maintains an equilibrium balance between the size of each command window and the amount of free resources.

19. The method in claim 17, wherein said command window defines a limit on the number of outstanding storage requests that each server may maintain.

20. The method in claim 17, wherein said allocating initially allocates a minimum sized command window.

21. The method in claim 17, wherein said minimum sized command window has a size no greater than the number of said resources divided by the number of said servers.

22. The method in claim 17, further comprising:

stopping said increasing processes when the amount of free resources falls below a lower limit;

restarting said increasing processes when the amount of said free resources rises above said lower limit; and

repeating said stopping and said restarting processes.

23. The method in claim 17, further comprising:

stopping said increasing processes for a server when said server stops utilizing said resources;

restarting said increasing processes when said server again begins utilizing said resources; and

repeating said stopping and said restarting processes.

24. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform a method for managing resources in a storage system being accessed by servers, said method comprising:

allocating a command window to each of said servers accessing said resources;

initially operating in an equilibrium achievement mode, wherein, while operating in said equilibrium achievement mode, said method comprises increasing the size of each server'"'"'s command window at a first rate as each of said servers completes a successful storage operation;

changing to a equilibrium maintenance mode when a storage request is unsuccessful, wherein, while operating in said equilibrium maintenance mode, said method comprises;

decreasing the size of each command window by a reduction factor; and

increasing the size of each server'"'"'s command window at a second rate less than said first rate as each of said servers completes a successful storage operation;

reverting to said equilibrium achievement mode when the level of free resources rises above an upper limit; and

repeating said changing and reverting processes.

25. The method in claim 24, wherein said method further comprises repeating process acquires and maintains an equilibrium balance between the size of each command window and the amount of free resources.

26. The method in claim 24, wherein said method further comprises command window defines a limit on the number of outstanding storage requests that each server may maintain.

27. The method in claim 24, wherein said method further comprises allocating initially allocates a minimum sized command window.

28. The method in claim 24, further comprising:

stopping said increasing processes when the amount of free resources falls below a lower limit;

restarting said increasing processes when the amount of said free resources rises above said lower limit; and

repeating said stopping and said restarting processes.

29. The method in claim 24, further comprising:

stopping said increasing processes for a server when said server stops utilizing said resources;

restarting said increasing processes when said server again begins utilizing said resources; and

repeating said stopping and said restarting processes.