System and method for reduced latency data transfers from flash memory to host by utilizing concurrent transfers into RAM buffer memory and FIFO host interface

• US 8,341,311 B1
• Filed: 11/18/2009
• Issued: 12/25/2012
• Est. Priority Date: 11/18/2008
• Status: Expired due to Fees

• Current Assignee: Entorian Technologies
• Original Assignee: Entorian Technologies
• Inventors: Szewerenko, Leland, Branca, Thomas
• Primary Examiner(s): Kim, Hong
• Application Number:
  US12/621,486
• Time in Patent Office: 1,133 Days
• Field of Search: 710/22, 710/38, 710/52, 711/103, 711/154
• US Class Current: 710/22
• CPC Class Codes: G06F 13/1673 using buffers
• CPC Class Codes: G06F 13/28 using burst mode transfer, e.g. direct memory access DMA, cycle steal G06F13/32 takes precedence

Abstract

A flash memory system having the capability of streaming data directly from flash memory to the interface of a host computer in order to substantially reduce latency of to-host transfers, while also maintaining the capabilities for caching and overlapped flash I/O provided by RAM DMA transfers. When data is read from the flash memory, the data is transferred into the RAM buffer and at the option of the memory controller, directly (via an intermediate FIFO) to the host interface. This results in a desirable reduction in the latency of data transfer because as soon as the first byte of data is read from the flash memory by the DMA engine, the data will be transferred directly to the host interface. Because the data is also being transferred to the buffer RAM, preferred embodiments of the present invention still provide the advantages of using an intermediate transfer buffer.

Claims

1. A flash memory device comprising:

a first flash memory for storing non-volatile data for a host;

a host interface for transferring data to the host;

a random-access memory (RAM) for storing data from the flash memory that are to be transferred to the host;

a first-in-first-out memory (FIFO) for receiving data from the flash memory;

a first flash direct-memory access (DMA) engine for receiving and transferring data among the first flash memory, the RAM, and the FIFO, the first flash DMA engine operably connected to the first flash memory via a first flash channel and the flash DMA engine being programmed to transfer data from the first flash memory to the host by transferring data from the first flash memory to the FIFO while simultaneously transferring the data from the first flash memory to RAM; and

a host interface DMA engine for receiving and transferring data among the host interface, RAM, and the FIFO, the host interface DMA engine being programmed to transfer data from the FIFO to the host interface as soon as data is transferred to the FIFO by the first DMA engine and to transfer data from the RAM once the transfer of data from the FIFO has been completed.

2. The device of claim 1 in which the RAM comprises multiple, separately accessible page buffers and in which transferring data from the first flash memory to RAM comprises transferring data from the first flash memory to a first RAM page buffer.

3. The device of claim 2 further comprising:

a second flash memory for storing non-volatile data for a host;

a second flash direct-memory access (DMA) engine for receiving and transferring data among the second flash memory and the RAM, the second flash DMA engine operably connected to the second flash memory via a second flash channel and the second flash DMA engine being programmed to transfer data from the second flash memory to a second RAM page buffer while the first DMA engine is transferring data from the first flash memory.

4. The device of claim 3 in which the host interface DMA engine is capable of assembling a data transfer to the host interface from a sequence of data transfers from the FIFO and the multiple RAM page buffers.

5. The device of claim 3 further comprising:

one or more additional flash memories for storing non-volatile data for a host;

one or more additional direct-memory access (DMA) engines for receiving and transferring data among the one or more additional flash memories and one or more additional RAM page buffers, the one or more additional flash DMA engines operably connected to corresponding one or more additional flash memories via one or more additional flash channels and the one or more additional flash DMA engines being programmed to transfer data from the one or more additional flash memories to one or more additional RAM page buffers while the first DMA engine is transferring data from the first flash memory to the FIFO and to the first RAM page buffer.

6. The device of claim 3 in which the data transferred by the first flash DMA engine and the data transferred by the second DMA engine are sequential portions of one block of data from flash memory.

7. The device of claim 3 in which the data transferred by the first flash DMA engine is a first portion of the data and the data transferred by the second DMA engine is a second portion of the data.

8. The device of claim 7 in which the first portion of data is the data needed immediately by the host and the second portion is sequential data.

9. The device of claim 1 in which transmitting the data through a first-in-first-out memory comprises streaming the data through the FIFO so that only a part of the first portion of the data is stored in the FIFO at any time.

10. The device of claim 1 in which the flash DMA engine is capable of direct transfer of contiguous memory from the flash memory and also the capability of gathering scattered non-contiguous memory that has been mapped as contiguous memory space in the flash memory.

11. The device of claim 1 further comprising an error connection circuit for detecting errors in data transferred from flash memory storage before the data is transferred to the FIFO or RAM.

12. The device of claim 1 in which the host interface comprises a SATA interface.

13. The device of claim 1 in which the flash memory comprises NAND flash memory, NOR flash memory, programmable read-only memory (PROM), electrically-erasable programmable read-only memory (EE-PROM), or read-only memory (ROM).

14. A nonvolatile memory system comprising:

a plurality of nonvolatile memory devices comprising a plurality of nonvolatile memory devices each connected to the system via a different corresponding flash channel;

a host interface;

a buffer memory; and

a circuit comprising, for each flash channel, a dedicated flash DMA engine for transferring data from each nonvolatile memory device connected to said flash channel;

wherein said host interface is coupled to said circuit and is enabled to receive data from the circuit and transfer the data to a host,

wherein said buffer memory is coupled to said circuit and is enabled to receive data from the circuit and transfer the data to a host,

wherein said circuit is coupled to each of the plurality of nonvolatile memory devices and is enabled to;

(i) stream data from a nonvolatile memory device to the host interface via a first-in-first out memory;

(ii) concurrently transfer data from a nonvolatile memory device to the buffer memory; and

(iii) transfer data from buffer memory to the host interface when the streaming of data from the nonvolatile memory device to the host interface has been completed;

wherein the circuit is coupled to each of the plurality of nonvolatile memory devices via the corresponding flash channels and the circuit is enabled to stream a first portion of sequential data from a nonvolatile memory device on a first flash channel to the host interface via a first-in-first out memory;

concurrently transfer a second portion of data from a nonvolatile memory device on a second flash channel to the buffer memory; and

transfer data from buffer memory to the host interface when the streaming of the first portion of data from the nonvolatile memory device to the host interface has been completed.

15. A method for accessing flash memory from one or more flash devices by a host computer, comprising:

concurrently directing a stream of data output from one or more flash devices to both a buffer memory and directly to a host interface comprises;

identifying the location of the desired data stored in flash memory;

reading the data from flash memory;

sending the data directly to a host interface without writing the data into a random-access memory (RAM) buffer and simultaneously writing the data to a RAM buffer;

after the data has been sent directly to a host interface, making the data stored in the RAM buffer available for transfer to the host interface;

wherein identifying the location of the desired data stored in flash memory comprises identifying the location of a first portion of sequential data stored in flash memory;

identifying the location of a second sequential portion of desired data stored in flash memory;

reading the second portion of data from flash memory;

writing the data to a second RAM buffer;

after the first portion of data has been sent directly to a host interface, making the second portion of data stored in the second RAM buffer available for transfer to the host interface; and

wherein the first portion of data is stored on flash memory connected to a first flash channel and the second portion of data is stored on flash memory connected to a second flash channel, and in which sending the data directly to a host interface without writing the data into a random-access memory (RAM) buffer comprises;

providing a first DMA engine connected to the first flash channel, the first DMA engine configured to transfer the first portion of data to a first-in-first out memory;

providing a second flash DMA engine connected to a second flash channel, the second DMA engine configured to transfer the second portion of data to a second RAM buffer;

providing a host interface DMA engine, the host interface DMA engine configured to either transfer data from the FIFO to the host interface or to transfer data from the second RAM buffer.