High performance integrated cached storage device

• US 5,890,207 A
• Filed: 11/27/1996
• Issued: 03/30/1999
• Est. Priority Date: 11/27/1996
• Status: Expired due to Term

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: Scaringella, Stephen Lawrence, Sne, Gal, Tung, Victor Wai Ner
• Primary Examiner(s): Chan, Eddie P.
• Assistant Examiner(s): KIM, HONG CHONG
• Application Number:
  US08/757,226
• Time in Patent Office: 853 Days
• Field of Search: 711/113, 711/114, 711/140, 711/169, 395/182.04, 395/185.07, 395/200.31
• US Class Current: 711/113
• CPC Class Codes: G06F 12/0866 for peripheral storage systems, e.g. disk cache
• CPC Class Codes: G06F 3/0619 in relation to data integrity, e.g. data losses, bit errors
• CPC Class Codes: G06F 3/0658 Controller construction arrangements
• CPC Class Codes: G06F 3/0689 Disk arrays, e.g. RAID, JBOD

Abstract

An integrated cached disk array includes host to global memory (front end) and global memory to disk array (back end) interfaces implemented with dual control processors configured to share substantial resources. Each control processor is responsible for 2 pipelines and respective Direct Multiple Access (DMA) and Direct Single Access (DSA) pipelines, for Global Memory access. Each processor has its own Memory Data Register (MDR) to support DMA/DSA activity. The dual processors each access independent control store RAM, but run the same processor independent control program using an implementation that makes the hardware appear identical from both the X and Y processor sides. Pipelines are extended to add greater depth by incorporating a prefetch mechanism that permits write data to be put out to transceivers awaiting bus access, while two full buffers of assembled memory data are stored in Dual Port Ram and memory data words are assembled in pipeline gate arrays for passing to DPR. Data prefetch mechanisms are included whereby data is made available to the bus going from Global Memory on read operations, prior to the bus being available for an actual data transfer. Two full buffers of read data are transferred from Global Memory and stored in DPR while data words are disassembled in the pipeline gate array, independent of host activity. Timing of system operations is implemented such that there is overlap of backplane requests to transfer data to/from memory, memory selection and data transfer functionality.

Claims

1. An integrated cache storage system moving host data in accordance with a host I/O protocol between a host and at least one array of storage devices for storage of said host data in and retrieval of said host data from said at least one array of storage devices, comprising:

at least one front end adapter having a first front end control processor and a second front end control processor and at least a first front end pipeline and a second front end pipeline for transferring said host data under control of said first front end control processor and said second front end control processor respectively, each of said first and second front end pipeline including, an I/O portion receiving a portion of said host data from said host and converting said host data from said host I/O protocol into at least one memory word, said at least one front end adapter further including a front end buffer memory receiving said at least one memory word;

a cache memory configured to receive said at least one memory word from said front end buffer memory to temporarily store said at least one memory word; and

at least one back end adapter having a first back end control processor and a second back end control processor and at least a first back end pipeline and a second back end pipeline for transferring said at least one memory word under control of said first back end control processor and said second back end control processor respectively, each of said first and second back end pipeline including, a back end buffer memory receiving said at least one memory word from said cache memory and an I/O portion receiving said memory word from said back end buffer memory and converting said at least one memory word into host data for transfer to said host according to said host I/O protocol.

2. The integrated cache storage system according to claim 1 wherein each of said first front end pipeline, said second front end pipeline, said first back end pipeline and said second back end pipeline further includes Error Detection and Correction circuitry between respective ones of said I/O portion and said buffer memory.

3. The integrated cache storage system according to claim 1 wherein said front end buffer memory comprises a first buffer memory section and a second buffer memory section and wherein said at least one memory word comprises a plurality of memory words and one of said first buffer memory section and said second buffer memory section is configured to receive a portion of said plurality of memory words from said I/O portion while the other of said first buffer memory section and said second buffer memory section is configured to transfer another portion of said plurality of memory words to said cache memory.

4. The integrated cache storage system according to claim 1 wherein said back end buffer memory comprises a first buffer memory section and a second buffer memory section and wherein said at least one memory word comprises a plurality of memory words and one of said first buffer memory section and said second buffer memory section is configured to receive a portion of said plurality of memory words from said cache memory while the other of said first buffer memory section and said second buffer memory section is configured to transfer another portion of said plurality of memory words to said I/O portion.

5. The integrated cache storage system according to claim 1 wherein said I/O portion of said at least one front end adapter converts data in accordance with one of a SCSI, Bus and Tag, and ESCON I/O protocol.

6. The integrated cache storage system according to claim 1 wherein said I/O portion of said at least one back end adapter converts data in accordance with a SCSI I/O protocol.

7. The integrated cache storage system according to claim 3 further including a bus transceiver configured to receive at least one memory word from one of said first memory buffer section and said second memory buffer section and to transfer said at least one memory word to a bus connected to said cache memory, and wherein said I/O portion of said front end adapter includes at least one memory word assembly register configured to transfer an assembled memory word to one of said first memory buffer section and said second memory buffer section:

wherein said system includes timing mechanisms that effect storage of at least one memory word in said bus transceiver available for transfer on said bus connected to said cache memory substantially simultaneous with loading of said at least one memory word in one of said first memory buffer section and said second memory buffer section while host data is received, converted and stored in said at least one memory word assembly register.

8. The integrated cache storage system according to claim 1 wherein said a least one front end adapter further includes first and second control stores storing a first control program for said first front end control processor and a second control program for said second front end control processor, said first control program and said second control program being substantially identical and specifying at least one memory address for accessing during execution, and wherein said at least one front end adapter includes a shared program memory having a first portion accessible to said first front end control processor and a second portion accessible to said second front end control processor, said first portion and said second portion having differing respective first and second address ranges, said at least one front end adapter having a transparency mechanism to receive said at least one memory address and determine which one of said first and second address ranges said at least one memory address is located in, and to determine a corresponding memory address in one of said first and second address ranges in which said at least one memory address is not located.

9. The integrated cache storage system according to claim 1 wherein said at least one front end adapter is configured to transfer said portion of host data from said host and to convert said portion of host data into said at least one memory word stored in said buffer memory. and wherein said at least one front end adapter is configured to write said at least one memory word in said cache memory in each of a first address and a second address.

10. A controller transferring data from a source to a destination, comprising:

a first control processor controlling at least a first pipeline selectable to transfer at least a first portion of said data from said source to said destination;

a second control processor controlling at least a second pipeline selectable to transfer at least a second portion of said data from said source to said destination;

a control program run by both said first control processor and said second control processor to control said first pipeline and said second pipeline to transfer said at least said first portion and said at least said second portion of data from said source to said destination, said control program specifying at least one memory address for accessing during execution;

shared resource memory having a first portion accessible to said first control processor and a second portion accessible to said second control processor, said first portion and said second portion having different respective first and second address ranges; and

a transparency mechanism receiving said at least one memory address and determining which one of said first address ranges said at least one memory address is located in for access by said first control processor, and determining a corresponding memory address in said second address range in which said at least one memory address is not located for access by said second control processor.

11. The controller of claim 10 electrically connected to a cache memory for receiving said data after leaving said source and prior to arriving at said destination, said cache memory configured to receive data from and transfer data to said first pipeline and said second pipeline, wherein said first and second pipelines each include an I/O portion receiving host data from a host connected to said controller said host data being in accordance with a host I/O protocol, said I/O portion converting said host data into at least one memory word, said controller further including a buffer memory receiving said at least one memory word from said I/O portion configured to transfer said at least one memory word to said cache memory.

12. The controller of claim 11 wherein said buffer memory comprises a first buffer memory section and a second buffer memory section and wherein said at least one memory word comprises a plurality of memory words and one of said first buffer memory section and said second buffer memory section is configured to receive a portion of said plurality of memory words from said I/O portion while the other of said first buffer memory section and said second buffer memory section is configured to transfer another portion of said plurality of memory words to said cache memory.

13. The controller of claim 11 wherein said I/O portion converts data in accordance with one of a SCSI, Bus and Tag, and ESCON I/O protocol.

14. The controller of claim 10 electrically connected to a cache memory storing at least one memory word and said first pipeline and said second pipeline are configured to receive said at least one memory word from said cache memory, wherein each of said first and second pipelines include, a buffer memory configured to receive said at least one memory word from said cache memory and an I/O portion receiving said at least one memory word from said buffer memory and converting said at least one memory word into host data for transfer to a host according to a host I/O protocol.

15. The controller of claim 14 wherein said buffer memory comprises a first buffer memory section and a second buffer memory section and wherein said at least one memory word comprises a plurality of memory words and one of said first buffer memory section and said second buffer memory section is configured to receive a portion of said plurality of memory words from said cache memory while the other of said first buffer memory section and said second buffer memory section is configured to transfer another portion of said plurality of memory words to said I/O portion.

16. The controller of claim 10 wherein said I/O portion converts data in accordance with a SCSI I/O protocol.

17. The controller of claim 12 further including a bus transceiver configured to receive said portion of said plurality of memory words from one of said first memory buffer section and said second memory buffer section and to transfer said at least one memory word to a bus connected to said cache memory, and wherein said I/O portion includes at least one memory word assembly register configured to transfer an assembled memory word to one of said first memory buffer section and said second memory buffer section,

wherein said system includes timing mechanisms that effect storage of at least one memory word in said bus transceiver available for transfer on said bus connected to said cache memory substantially simultaneous with loading of said at least one memory word in one of said first memory buffer section and said second memory buffer section while data is received for conversion into a memory word and stored in said at least one memory word assembly register.

18. The controller of claim 10 wherein said controller is implemented in an integrated cache disk array system, said source is a host computer and said data is host data transferred in accordance with a host I/O protocol, and said destination is at least one array of disk drives for storage of said host data in and retrieval of said host data from said at least one array of disk drives.

19. The controller of claim 12 wherein said controller is configured to transfer said host data from said host computer and to convert said host data into said at least one memory word stored in said buffer memory, and wherein said at least one controller is configured to write said at least one memory word in said cache memory in each of a first address and a second address.

20. A method of pipelining data for transfer between a data source and a data destination, said method comprising the steps of:

receiving data from said data source during a write operation and assembling said data into a plurality of data destination words;

storing each of said plurality of data destination words individually in at least one register pipeline stage;

configuring a first and a second buffer memory, said first and said second buffer memory being in communication with said at least one register pipeline stage to receive respective ones of each of said plurality of data destination word therefrom;

transferring a first write portion of said plurality of data destination words into said first buffer memory until a first selected number of data destination words are transferred to said first buffer memory;

transferring said first write portion of said plurality of data destination words from said first buffer memory to said data destination until said first selected number of data destination words are transferred from said first buffer memory to said data destination;

transferring a second write portion of said plurality of data destination words assembled and stored in said at least one register pipeline stage into said second buffer memory until a second selected number of data destination words are transferred to said second buffer memory, while said step of transferring said first write portion of said plurality of data destination words from said first buffer memory to said data destination is taking place; and

transferring said second write portion of said plurality of data destination words from said second buffer memory to said data destination until said second selected number of data destination words are transferred from said second buffer memory to said data destination, while said first buffer memory is available to receive another write portion of said plurality of data destination words assembled for transfer by said at least one register pipeline stage.

21. The method of claim 20 wherein said at least one register pipeline stage comprises a first register pipeline stage and a second register pipeline stage, said first register pipeline stage is configured to receive said data and assemble said data into respective ones of said plurality of data destination words, and said second register pipeline stage is configured to receive said respective ones of said plurality of data destination words from said first register pipeline stage and to transfer said respective ones of said plurality of data destination words to one of said first buffer memory and said second buffer memory.

22. The method of claim 20 wherein said first and second buffer memory are configured using dual port random access memory (DPR) and wherein a first port of said DPR is configured for handling data transferred between said DPR and said data source, and a second port is configured for handling data transferred between said DPR and said data destination.

23. The method of claim 20 wherein each of said plurality of data destination words is processed by an error detection and correction mechanism prior to being received by said first buffer memory and said second buffer memory.

24. The method of claim 20 wherein each of said plurality of data destination words comprises 64 bits of data and 8 Error Correction Code bits and said first selected number of data destination words and said second selected number of data destination words each equal 32.

25. The method of claim 20 wherein said steps of transferring said first portion of said plurality of data destination words from said first buffer memory to said data destination and transferring said second portion of said plurality of data destination words from said second buffer memory to said data destination include passing each of said plurality of data destination words from one of said first buffer memory and said second buffer memory to a data transceiver for holding until said data destination is available to receive each of said plurality of data destination words.

26. The method of claim 20 wherein said data source is a computer system and wherein said data destination is a cache memory in an integrated cache disk array.

27. The method of claim 20 further including the steps of:

receiving each of a plurality of data destination words from said data destination during a read operation;

transferring a first read portion of said plurality of data destination words into said first buffer memory until said first selected number of data destination words are transferred to said first buffer memory;

storing each of said first read portion of said plurality of data destination words individually in said at least one register pipeline stage, disassembling each individual data destination word into data and passing said data to said data source;

transferring a second read portion of said plurality of data destination words into said second buffer memory until said second selected number of data destination words are transferred to said second buffer memory, while said step of storing each of said first read portion of said plurality of data destination words individually in said at least one register pipeline stage is taking place; and

storing each of said second read portion of said plurality of data destination words from said second buffer memory individually into said at least one register pipeline stage, disassembling each individual data destination word into data and passing said data to said data source, while said first buffer memory is available to receive another portion of said plurality of data destination words from said data destination.

28. The method of claim 27 wherein said at least one register pipeline stage comprises a first register pipeline stage and a second register pipeline stage, said second register pipeline stage is configured to receive said destination data words from said first buffer memory and said second buffer memory and to pass said destination data words to said first pipeline stage, and said first pipeline stage is configured to disassemble said data destination words into data and to transfer said data to said data source.

29. The method of claim 27 wherein each of said plurality of data destination words is processed by an error detection and correction mechanism prior to being received by said at least one register pipeline stage.