Detection of sequential data stream for improvements in cache data storage

• US 4,468,730 A
• Filed: 11/27/1981
• Issued: 08/28/1984
• Est. Priority Date: 11/27/1981
• Status: Expired due to Term

• Current Assignee: Storage Technologies Corporation (Oracle Corporation)
• Original Assignee: Storage Technologies Corporation (Oracle Corporation)
• Inventors: Coulson, Richard L., Blickenstaff, Ronald L., Trede, Brian E., Dodd, P. David, Moreno, Robert J.
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): Watson, Mark P.
• Application Number:
  US06/325,350
• Time in Patent Office: 1,005 Days
• Field of Search: 364/200 MS File, 364/900 MS File
• US Class Current: 711/113
• CPC Class Codes: G06F 12/0862 with prefetch
• CPC Class Codes: G06F 12/0866 for peripheral storage systems, e.g. disk cache
• CPC Class Codes: G06F 2212/6026 Prefetching based on access pattern detection, e.g. stride based prefetch
• CPC Class Codes: G06F 2212/6028 Prefetching based on hints or prefetch instructions

Abstract

A method for detection of a sequential data stream which can be performed without host computer intervention is disclosed featuring examination of a data record and channel program during read operations for signals indicative that the data is not part of a sequential data stream, for example, embedded seek instructions. If a particular sought for record does not contain such indications, the successive record or records may then be staged to a faster access memory device such as a solid-state cache. The invention is described in a plug-compatible, software-transparent configuration.

Claims

1. A method for increasing efficiency of operation of a computer system of the class comprising:

a host computer including an arithmetic and logic unit and main memory means; and

a data storage unit comprising a long-term magnetic data storage means, a solid-state caching memory means and a cache manager means,

said data storage unit being connected to said host computer by a channel, said channel comprising means for communication of commands and data between said host computer and said data storage unit, the commands output by the host to the data storage unit in calling for data not including an indication of whether data called-for is part of a sequence of such requests;

said method comprising the following steps, performed by said data storage unit solely under control of said cache manager, and external to said channel and said host computer;

examining the data called for by the host from the long-term magnetic memory means;

determining if the called-for data contains indicators indicating that the called-for data is not part of a sequential data stream;

staging data not called for by said host computer to said cache memory means upon determination that said called-for data does not contain said indicators, said not called-for data staged to said cache memory being stored on said long-term magnetic storage media in a predetermined physical relationship to said called-for data; and

avoiding staging data to said cache memory means upon determination that said called-for data contains indicators that the called-for data is not part of a sequential data stream.

2. The method of claim 1 wherein said indicators comprise commands contained within said called-for data directing additional data to be read from locations other than those disposed in said predetermined physical relationship to said called-for data on said long-term magnetic storage media.

3. The method of claim 1 wherein said data is stored on disk drive media organized into logical tracks and said staging is only performed when said indicators are not present and when said previously called-for data includes the starting portion of one of said logical tracks.

4. The apparatus of claim 1 wherein said long-term magnetic data storage means comprise disk drive means wherein data is stored on concentric tracks on disks, and said staging is performed with respect to an entire track at a time, successive tracks being staged to said cache as previously staged tracks are called for by said host.

5. Method for improving the efficiency of a data processing system of the type comprising a host computer connected via channel means to a memory system comprising a solid-state cache memory means and long-term data storage means, said host being adapted to issue commands calling for specific data records stored on said long-term storage means, said commands not including an indication of whether a particular request is one of a group of requests for data stored sequentially on said long-term storage means, comprising the steps of:

examining each individual host originated command calling for a particular data record and the contents of the solid-state cache memory means to determine whether the data thus called for is stored in said solid-state memory means external to said channel, and, if said data is stored in said solid-state memory means, transmitting said data to said host;

if said data is not stored in said solid-state memory means, accessing said data on said long-term data storage means and passing it over said channel means to said host, and examining said data passed from said long-term data storage means to said host in order to determine whether said called-for record is one of a sequence of such records; and

if said determination indicates that said called-for record may be one of such a sequence, determining the next record in said sequence and reading said next record from said long-term storage means into said solid-state memory means in anticipation of said next record being subsequently called for by said host.

6. The method of claim 5 wherein the determination that a given record is one of a sequence is accomplished by examination of said record for indicators signaling that said data is not part of such a sequence.

7. The method of claim 6 wherein said long-term data storage means comprises magnetic disk drive means, said disks being organized into tracks, said indicators signaling that a given record is not part of a sequence of such records comprising commands instructing that records not on the same track of the same disk be supplied to said host.

8. The method of claim 5 wherein said records are stored on long-term storage means, said long-term storage means comprising disk drive means adapted to store data organized into tracks, said data being read from said long-term storage means into said solid-state memory means track by track.

9. A method of control of storage of digital data in storage means comprising:

long-term magnetic storage means;

faster access solid-state cache memory means; and

controller means;

said method comprising the steps of;

examining the sequence of commands output by a host computer in order to access data required, said commands not including an indication of whether said data required is part of a sequential data stream, and the data accessed by the host computer, for indicators that said data is not part of a sequential data stream;

staging a quantity of additional data to said cache memory means where said indicators are not present, in anticipation of said additional data being subsequently called for by said host computer; and

avoiding staging additional data to said cache means where said indicators are present.

10. The method of claim 9 wherein said examining step comprises the step of determining whether said sequence of commands includes instructions to read other data not stored sequentially on said long-term storage means with respect to said data required.

11. Apparatus for storage and retrieval of digital data for connection to a host computer of the type outputting commands for retrieval of data, said commands not including explicit indication of whether a particular data request is one of a sequence of requests for sequential portions of a given data file, said apparatus comprising:

long-term data storage means;

solid-state cache memory means; and

cache controller means;

said cache controller means being adapted to examine commands output by said host in effectuating retrieval of data, and to examine said data for embedded commands, said controller being adapted to determine whether said commands indicate that a particular data request is likely not to be part of a sequence of such requests, and to stage additional data from said long-term memory means to said cache memory means if no such indicators are present in the processing of a particular data request, in anticipation of further requests for sequential portions of a given data record.

12. The apparatus of claim 11 wherein said long-term data storage means comprises magnetic disk drive storage means wherein said data stored on said disks is organized into logical tracks and said data is staged from said disk drives to said cache memory means in multiples of one track.