Method for accessing one or more streams in a video storage system using multiple queues and maintaining continuity thereof

• US 5,802,394 A
• Filed: 06/06/1994
• Issued: 09/01/1998
• Est. Priority Date: 06/06/1994
• Status: Expired due to Fees

• Current Assignee: Starlight Networks, Inc. (HP Inc.)
• Original Assignee: Starlight Networks, Inc. (HP Inc.)
• Inventors: Baird, Randall B., McFadden, Martin J.
• Primary Examiner(s): Harvey, Jack B.
• Assistant Examiner(s): Lefkowitz, Sumati
• Application Number:
  US08/254,576
• Time in Patent Office: 1,548 Days
• Field of Search: 395/438-441, 395/859, 395/825-826
• US Class Current: 710/5
• CPC Class Codes: G06F 3/0613 in relation to throughput
• CPC Class Codes: G06F 3/0659 Command handling arrangements, e.g. command buffers, queues, command scheduling
• CPC Class Codes: G06F 3/0689 Disk arrays, e.g. RAID, JBOD
• CPC Class Codes: G11B 20/1217 on discs
• CPC Class Codes: G11B 27/105 of operating discs
• CPC Class Codes: H04N 21/21 Server components or server architectures
• CPC Class Codes: H04N 21/23 Processing of content or additional data; Elementary server operations; Server middleware

Abstract

A method is disclosed for interfacing a video storage system, such as a disk array, and a video stream application executing on a computer network or stand-alone system. The interface provides a method for storing video stream data by allocating storage space in units of stripes and for accessing the data in units of fixed length segments, wherein each stripe includes one fixed length segment from each disk of the array. The interface also queues submitted I/O transaction requests so that I/O transactions requested on a current I/O cycle i are not performed until the next I/O cycle i+1. The I/O transactions thus queued are performed in an order which permits a maximum number of I/O transactions to be performed each I/O cycle, regardless of the order in which they were submitted. In addition, the interface also queues the submitted I/O transaction requests so that the completion of an I/O transaction on a current I/O cycle i is not returned until the next I/O cycle i+1. The completions of the I/O transactions thus queued are returned in the order they were submitted regardless of the order in which they were performed. Furthermore, the interface provides a state machine which controls the amount of data accessed each I/O cycle for a corresponding stream. The state machine permits no more than a particular allotted amount of data to be accessed each I/O cycle, wherein the allotted amount of data is selected according to the bandwidth requirement of the corresponding stream.

Claims

1. A method for accessing data of one or more streams in a disk array of at least one disk comprising the steps of:

(a) enqueuing one or more command descriptors for performing I/O transactions, corresponding to each request that issued during a current cycle i, into an inactive periodic queue, during said cycle i,

(b) delaying performing said I/O transactions of said command descriptors in said inactive periodic queue until a subsequent cycle i+1,

(c) at a beginning of said cycle i+1, placing said command descriptors in said inactive periodic queue into an active periodic queue,

wherein during each cycle, I/O transactions are performed for each command descriptor in said active periodic queue, but not performed for command descriptors in said inactive periodic queue.

2. The method of claim 1 further comprising, prior to step (a), the steps of:

(d) during said cycle i, enqueuing an I/O descriptor for each I/O transaction requested during said cycle i into a submit pending queue,

(e) at said beginning of said cycle i+1;

(e1) dequeuing each of said I/O descriptors from said submit pending queue and enqueuing each dequeued I/O descriptor into a submitted I/O transaction queue associated with the stream on which said I/O transaction of said I/O descriptor corresponds,

(e2) for each of said I/O descriptors, performing steps (a)-(c).

3. The method of claim 2 further comprising, prior to said step (d), the steps of:

(f) opening a chunk assigned to a particular stream on which one or more I/O transactions are to be performed and allocating resources for said chunk.

4. The method of claim 3 further comprising the steps of:

(f1) allocating bandwidth for transmitting data read or written during each I/O transaction performed on said particular stream,

(f2) generating an open descriptor for said particular stream including a submitted I/O transaction queue for storing I/O descriptors of submitted pending I/O transaction requests, and

(f3) generating a completion context in which command descriptors for completed I/O transactions are enqueued.

5. The method of claim 4 further comprising the steps of:

(g) if said chunk is closed but a second chunk is specified to be opened immediately thereafter, assigning said resources allocated to said chunk to said second chunk.

6. The method of claim 1 further comprising the steps of:

(d) prior to said cycle i+1, ordering each command descriptor in said inactive periodic queue to increase the number of I/O transactions that can be performed during said cycle i+1, and

(e) during said cycle i+1;

(e1) dequeuing up to a predetermined number of command descriptors from the head of said active periodic queue and enqueuing each of said dequeued command descriptors into an in-progress queue, and

(e2) sequentially issuing to said disk array a command for performing said I/O transaction corresponding to each command descriptor in said in-progress queue.

7. The method of claim 1 further comprising the steps of, as each I/O transaction completes:

(d) dequeuing said command descriptor corresponding to a completed I/O transaction,

(e) enqueuing said dequeued command descriptor corresponding to said completed I/O transaction onto a completion context corresponding to a stream for which said I/O transaction was performed, and

(f) signaling a semaphore associated with said completion context onto which said command descriptor for said completed I/O transaction was dequeued.

8. The method of claim 6 further comprising the steps of:

(f) as each I/O transaction completes, dequeuing said command descriptor corresponding to said completed I/O transaction,

(g) if said active periodic queue is not empty, performing said steps (e), (e1) and (e2) and

(h) if said active periodic queue is empty, dequeuing from an aperiodic queue one or more command descriptors for performing I/O transactions other than for reading or writing stream data, enqueuing each command descriptor that was dequeued from said aperiodic queue into said in-progress queue and sequentially issuing to said disk array a command for performing said I/O transaction corresponding to each command descriptor in said in-progress queue.

9. The method of claim 2 wherein in said step (e1), each I/O descriptor is enqueued into said submitted I/O transaction queue in the same order in which said requests for performing said I/O transactions are requested.

10. The method of claim 9 comprising the steps of:

(f) dequeuing said I/O descriptors from said submitted I/O transaction queue in the same order in which they are stored when said I/O transactions corresponding to said I/O descriptors complete.

11. The method of claim 10 further comprising the step of:

(g) limiting the number of outstanding submitted requests for performing I/O transactions for each stream to a predetermined maximum number.

12. The method of claim 11 further comprising the steps of:

(h) decreasing a count of the number of outstanding submitted requests for performing I/O transactions on said particular stream for each I/O descriptor dequeued from said submitted I/O transaction queue of said particular stream.

13. The method of claim 6 further comprising the step of:

(f) if at the end of said cycle i+1, at least one I/O transaction corresponding to a command descriptors enqueued into said active periodic queue at said start of said cycle i+1 is not complete, adjusting a duration of said cycle i+2 to prevent drifting of boundaries of said cycles.

14. The method of claim 1 further comprising the step of:

(d) limiting the number of outstanding submitted requests for performing I/O transactions for each stream to a predetermined maximum number.

15. The method of claim 14 further comprising the steps of:

(e) dequeuing an I/O descriptor from a submitted I/O transaction queue of a particular stream when said I/O transaction corresponding to said I/O descriptor completes, and

(f) decreasing a count of the number of outstanding submitted requests for performing I/O transactions on said particular stream.

16. The method of claim 1 further comprising the step of:

predefining the cycle time duration as N_d S/W_base, where N_d is the number of disks in the array, S is the number of bits consumed or produced for each stream during the cycle and W_base is a base bandwidth of each stream.

17. A method for performing I/O transactions on stream oriented data comprising the steps of:

(a) for each request to perform an I/O transaction on a stream issued during a current cycle i, enqueuing an I/O descriptor into a submit pending queue,

(b) at the beginning of the next cycle i+1, dequeuing each of said I/O descriptors from said submit pending queue and enqueuing said dequeued I/O descriptors into a submitted I/O transaction queue associated with said stream on which said I/O transaction of said I/O descriptor is to be performed, said I/O descriptors being enqueued into each associated submitted I/O transaction queue in the same order in which said requests to perform said corresponding I/O transactions are requested,

(c) performing said I/O transactions in an order that is independent of said order in which said I/O requests corresponding to said I/O transactions are stored in said submitted I/O transaction queues, and

(d) maintaining continuity of said streams by retrieving said I/O descriptors from said respective submitted I/O transaction queue in the same order in which said I/O transactions are requested and independently of a specific order in which said I/O transactions corresponding to said I/O descriptors were performed.

18. The method of claim 17 wherein said step (c) further comprises the steps of:

(c1) enqueuing one or more command descriptors for performing I/O transactions, corresponding to each request that issued during said cycle i, into an inactive periodic queue, during said cycle i,

(c2) delaying performing said I/O transactions of said command descriptors in said inactive periodic queue until said cycle i+1,

(c) at a beginning of said cycle i+1, placing said command descriptors in said inactive periodic queue into an active periodic queue,

wherein during each cycle, I/O transactions are performed for each command descriptor in said active periodic queue, but not performed for command descriptors in said inactive periodic queue.

19. The method of claim 17 further comprising, prior to said step (a), the steps of:

(e) opening a chunk assigned to a particular stream on which one or more I/O transactions are to be performed and allocating resources for said chunk.

20. The method of claim 19 further comprising the steps of:

(e1) allocating bandwidth for transmitting data read or written during said I/O transaction,

(e2) generating an open descriptor for said stream including a queue for storing I/O descriptors for submitted pending I/O transaction requests, and

(e3) generating a completion context in which command descriptors for completed I/O transactions are enqueued.

21. The method of claim 20 further comprising the steps of:

(f) if said chunk is closed but a second chunk is specified to be opened immediately thereafter, assigning said resources allocated to said chunk to said second chunk.

22. The method of claim 18 further comprising the steps of:

(e) prior to said cycle i+1, ordering each command descriptor in said inactive periodic queue to increase the number of I/O transactions that can be performed during said cycle i+1, and

(f) during said cycle i+1;

(g1) dequeuing up to a predetermined number of command descriptors from the head of said active periodic queue and enqueuing each of said dequeued command descriptors into an in-progress queue, and

(g2) sequentially issuing to an array of at least one disk, a command for performing said I/O transaction corresponding to each command descriptor in said in-progress queue.

23. The method of claim 18 further comprising the steps of, as each I/O transaction completes:

(e) dequeuing said command descriptor corresponding to said completed I/O transaction,

(f) enqueuing said dequeued command descriptor corresponding to said completed I/O transaction onto a completion context corresponding to a stream for which said I/O transaction was performed, and

(g) signaling a semaphore associated with said completion context onto which said command descriptor for said completed I/O transaction was dequeued.

24. The method of claim 22 further comprising the steps of:

(h) as each I/O transaction completes, dequeuing said command descriptor corresponding to said completed I/O transaction,

(i) if said active periodic queue is not empty, performing said steps (g), (g1) and (g2) and

(j) if said active periodic queue is empty, dequeuing from an aperiodic queue one or more command descriptors for performing I/O transactions other than for reading or writing stream data, enqueuing each command descriptor that was dequeued from said aperiodic queue into said in-progress queue and sequentially issuing a command for performing said I/O transaction corresponding to each command descriptor in said in-progress queue.

25. The method of claim 22 further comprising the step of:

(h) if at the end of said cycle i+1, at least one I/O transaction corresponding to a command descriptors enqueued into said active periodic queue at said start of said cycle i+1 is not complete, adjusting a duration of said cycle i+2 to prevent drifting of boundaries of said cycles.

26. The method of claim 17 further comprising the step of:

(e) limiting the number of outstanding submitted requests for performing I/O transactions for each stream to a predetermined maximum number.

27. The method of claim 26 further comprising the steps of:

(f) dequeuing an I/O descriptor from a submitted I/O transaction queue of a particular stream when said I/O transaction corresponding to said I/O descriptor completes, and

(g) decreasing a count of the number of outstanding submitted requests for performing I/O transactions on said particular stream.

28. The method of claim 17 further comprising the step of:

predefining the cycle time duration as N_d S/W_base, where N_d is the number of disks in the array, S is the number of bits consumed or produced for each stream during the cycle and W_base is a base bandwidth of each stream.