Method for scheduling I/O transactions in a data storage system to maintain the continuity of a plurality of video streams

• US 5,734,925 A
• Filed: 09/10/1996
• Issued: 03/31/1998
• Est. Priority Date: 11/17/1992
• Status: Expired due to Fees

• Current Assignee: Starlight Networks, Inc. (Plantronics, Inc.)
• Original Assignee: Starlight Networks, Inc. (Plantronics, Inc.)
• Inventors: Baird, Randall B., Tobagi, Fouad A., Pang, Joseph W. M., Gang, Joseph M. Jr., McFadden, Martin J.
• Primary Examiner(s): Lee, Thomas C.
• Assistant Examiner(s): PERVEEN, REHANA
• Application Number:
  US08/711,050
• Time in Patent Office: 567 Days
• Field of Search: 395/826, 395/825, 395/841, 395/800
• US Class Current: 710/6
• CPC Class Codes: G06F 16/10 File systems; File servers
• CPC Class Codes: G06F 3/061 Improving I/O performance
• CPC Class Codes: G06F 3/0659 Command handling arrangements, e.g. command buffers, queues, command scheduling
• CPC Class Codes: G06F 3/067 Distributed or networked storage systems, e.g. storage area networks [SAN], network attached storage [NAS]
• CPC Class Codes: G11B 2220/20 Disc-shaped record carriers
• 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
• CPC Class Codes: H04N 7/17336 Handling of requests in head-ends
• CPC Class Codes: H04N 7/24 Systems for the transmission of television signals using pulse code modulation H04N21/00 takes precedence

Abstract

A method for operating a disk storage system, comprising a disk and forming part of a communications network, simultaneously maintains the continuity of a plurality of data streams. Typically, each stream transfers video data to or from the disk storage system. Illustratively, each of the data streams is produced in the network at a rate of W_base bits/sec and consumed by the disk storage system or produced by said disk storage system and consumed in the network at a rate of W_base bits/sec. One I/O transaction is performed for each stream in each of a plurality of I/O cycles of duration S/W_base, wherein in each I/O transaction a segment of S bits is retrieved from or stored in the disk. The number of streams whose continuity can be maintained in this manner is limited by the number of I/O'"'"'s which can be performed in a cycle of duration S/W_base. More generally, when a stream has a bit rate (A/B)W_base bits/sec where A and B are integers chosen independently for each stream, then for this stream A I/O transactions are performed in B cycles of duration S/W_base. If the number of disks is N_d,N_d ≧1, then the I/O cycles have a duration N_d S/W_base and then A I/O transactions are performed in each disk for each stream in B I/O cycles.

Claims

1. A method for scheduling I/O transactions for a data storage system comprising N_d storage units, said data storage system storing data for a plurality of full motion video streams, each of said streams having a bit rate of (A/B)W_base where A and B are integers chosen independently for each of said streams, said method comprising the steps of

maintaining the continuity of said full motion video streams simultaneously by performing for each of said streams A I/O transactions in each of said N_d storage units within each set of B I/O cycles in a sequence of sets of B I/O cycles, each I/O cycle having a duration of N_d S/W_base,

in each of said I/O transactions a segment of S bits is retrieved from one of said storage units or written into one of said storage units,

the duration of each I/O transaction being shorter than S/W_base,

for streams in which segments are retrieved from said storage system, the segments of a stream retrieved from said storage system in each particular set of B I/O cycles being consumed in a particular set of B consumption cycles in a sequence of sets of B consumption cycles, the particular set of B consumption cycles beginning after the first I/O transaction is performed for the stream in the particular set of B I/O cycles, the duration of a consumption cycle being N_d S/W_base,

for streams in which segments are written into said storage system, the segments of a stream written into said storage system in each particular set of B I/O cycles being produced in a particular set of B production cycles in a sequence of sets of B production cycles, the particular set of B production cycles ending prior to performing the last I/O transaction for the stream in the particular set of B I/O cycles;

the duration of a production cycle being N_d S/W_base.

2. The method of claim 1 wherein said storage system is located in a communications network.

3. The method of claim 1 wherein the network produces or consumes W_base bits/sec for each of the streams.

4. The method of claim 1 wherein said streams share buffer capacity in a local buffer.

5. The method of claim 1

wherein said streams are divided into G groups,

wherein the production or consumption cycles of the streams within each group are aligned with each other, but the production and consumption cycles of successive groups are offset from the production and consumption cycles of the previous group by (1/G)^th of an I/O cycle, and wherein each I/O cycle is divided into G I/O subcycles of duration N_d S/W_base G.

6. The method of claim 5 wherein in each I/O subcycle there is one I/O transaction performed in each storage unit for each stream in the group whose production or consumption cycle begins immediately after the subcycle.

7. The method of claim 5 wherein in each subcycle, one I/O transaction is performed for each stream in each of N_d groups.

8. The method of claim 1 wherein the number of I/O transactions performed in each I/O cycle for each storage unit for active streams does not exceed a predetermined number.

9. The method of claim 8 further comprising the step of activating one or more new streams if the number of I/O transactions performed in each I/O cycle for each storage unit for active streams is less than said predetermined number.

10. The method of claim 9 wherein excess capacity for I/O transactions in said storage units is used for start-up data for said newly activated streams.

11. The method of claim 9 wherein new streams are activated according to a predetermined service discipline.

12. The method of claim 11 wherein the predetermined service discipline is chosen from the group comprising first-come-first-served, round robin, last-come-first-served, and shortest job first.

13. The method of claim 10 wherein in each I/O cycle, the I/O transactions of previously active streams and I/O transactions for the start-up data of newly activated streams are sorted together.

14. The method of claim 10 wherein in each I/O cycle, the I/O transactions of previously active streams are sorted and performed in one subcycle and the I/O transactions for the start-up data of newly activated streams are separately sorted and performed in a separate subcycle.

15. The method of claim 1 wherein said network is a local area network.

16. The method of claim 1 wherein said streams contain video data.

17. The method of claim 1 wherein I/O transactions for said streams have a predetermined order which does not change from one I/O cycle to the next.

18. The method of claim 1 wherein the I/O transactions have an order which is separately determined for each I/O cycle.

19. The method of claim 1 wherein the storage system is a Redundant Array of Inexpensive Disks.

20. The method of claim 1 wherein B has a fixed value for all streams N_super, where N_super is an integer.

21. The method of claim 1 wherein segments of said streams are allocated to the storage units in said array in a regular cyclic pattern.

22. The method of claim 1 wherein for each stream the A I/O transactions for each storage unit are uniformly spread in B I/O cycles.

23. The method of claim 20 wherein N_super I/O cycles form a supercycle, wherein in each I/O cycle of a supercycle, for each stream L stripes are fetched from or written into said storage system, wherein each stripe comprises one segment of S bits fetched from or written into each of said N_d storage units, and wherein for each stream the variance of the number L over the cycles of a supercycle is minimized.

24. The method of claim 23 wherein the variance of the total number of stripes fetched from or written into said storage system, for all of said streams in each cycle of a supercycle is minimized.

25. The method of claim 1 wherein segments are grouped into data stripes of N_d consecutive segments, wherein said array is organized into bin stripes comprising locations in each of said N_d storage units, and wherein data stripes are randomly allocated to bin stripes.

26. A method for scheduling I/O transactions for a data storage system comprising N_d storage units to enable full motion video streams to be written into said data storage system and to be retrieved from said data storage system, said data storage system storing data for a plurality of said full motion video streams, said streams being divided into G groups, said method comprising the steps of:

maintaining simultaneously the continuity of said full motion video streams to be written into said data storage system and to be retrieved from said data storage system by performing for each of said streams a predetermined number of I/O transactions in each of said N_d data storage units in each of a plurality of successive I/O cycles, each I/O cycle having a duration of N_d S/W_base, where S is a number of bits and W_base is a basic rate,

in each of said I/O transactions a segment of S bits is retrieved from one of said storage units or written into one of said storage units,

the duration of an I/O transaction being shorter than the duration of an I/O cycle so that multiple I/O transactions can be scheduled for each said data storage unit in an I/O cycle,

for streams in which segments are retrieved from said storage system, the segments of a stream being retrieved from the storage system in each particular I/O cycle being consumed in a particular consumption cycle in a sequence of consumption cycles, the particular consumption cycle beginning after a first I/O transaction for the stream in the particular I/O cycle,

for streams in which segments are written into said storage system, all of the segments of a stream being written into the storage system in each particular I/O cycle being produced in a particular production cycle in a sequence of production cycles, the particular production cycle ending prior to a last I/O transaction for the stream in the particular I/O cycle, and

wherein the production or consumption cycles of the streams within each group are aligned with each other, but the production and consumption cycles of successive groups are offset from the production and consumption cycles of the previous group by (1/G)^th of an I/O cycle, and wherein each I/O cycle is divided into G I/O subcycles of duration N_d S/W_base G.