Memory transaction with reduced latency

Abstract

A solution for performing reduced latency memory read transactions is disclosed. In one example, a storage apparatus has a memory array that includes: a flash device having a data register, a memory interface coupled to the memory array and a buffer set that includes at least one buffer suitable for use as a prefetch buffer. The memory interface, in response to a memory read transaction request, performs a read operation and, if stored data exists within the memory array that meets a prefetch selection criterion, also performs an internal read operation. The internal read operation includes allocating a prefetch buffer in the buffer set and storing the data as prefetch data in the prefetch buffer. If the memory interface receives a second memory read transaction request for data that is currently available as prefetch data, the memory interface responds by performing a forwarding transaction that includes retrieving the prefetch data from the prefetch buffer and forwarding the prefetch data to a host.

Description

FIELD OF INVENTION

The present invention relates to solutions for achieving reduced latency memory read transactions in storage or memory devices. More particularly, these solutions achieve reduced latency memory read transactions by allocating a prefetch buffer from a set of buffers that includes at least one buffer, such as a data register, from a flash device.

BACKGROUND

The relentless increase in Internet usage and e-commerce as well as the proliferation of networked digital computing devices has driven the demand for not only large capacity data storage but also high data availability and sophisticated data management, such as storage virtualization. To meet these needs, many vendors offer networked attached storage (NAS) devices, storage array network (SAN) devices and sophisticated servers that have directly attached arrays of mass storage. These solutions can provide users with managed access to their data directly from their computing device or through a network and with high data availability. These solutions are not without challenges, however.

One challenge includes reducing latency in memory read transactions. For example, in enterprise database applications or in on-line transaction processing, massive amounts of data may need to be sorted or filtered. This requires that these applications have access to storage devices that can not only provide large data capacity but perform low latency memory read transactions for data stored randomly in the storage devices. Another example includes on-demand video applications. These applications must have access to storage devices that can retrieve large amounts of sequential data, such as video, with low latency and in multiple streams in order to maximize viewing enjoyment of the video in real-time by multiple end-users.

Yet another example requiring reduced latency in memory read transactions includes a storage virtualization application that can aggregate a pool of storage that can be used by requesting hosts. This pool of storage may include a set of storage devices that may be interconnected through a LAN, WAN or both, and that may use mass storage devices that have relatively high memory transaction read latency, such as hard disk drives (HDD), or that are coupled to a network with a relatively low network bandwidth.

Consequently, a need exists for a storage device that can provide low latency memory read transactions in a variety of applications.

SUMMARY

A solution for performing reduced latency memory read transactions is disclosed. In one example, this solution may include a storage apparatus having a memory array that includes: a flash device having a data register, a memory interface coupled to the memory array and a buffer set that includes at least one buffer that is suitable for use as a prefetch buffer. The buffer set, where one buffer in the buffer set includes a data register, is coupled to the memory interface. The memory interface, in response to a memory read transaction request, performs a read operation and, if stored data exists within the memory array that meets at least one prefetch selection criterion, also performs an internal read operation. The internal read operation includes allocating a prefetch buffer in the buffer set and storing the data as prefetch data in the prefetch buffer. If the memory interface receives a second memory read transaction request for data and if the data is currently available as prefetch data in the prefetch buffer, the memory interface responds by performing a forwarding transaction that includes retrieving the prefetch data from the prefetch buffer and forwarding the prefetch data to a host.

BRIEF DESCRIPTION OF DRAWINGS

FIG. 1 is a block diagram of a reduced latency storage apparatus in accordance with a first embodiment of the present invention.

FIG. 2 is a block diagram of a reduced latency memory interface for use in a storage apparatus in accordance with a second embodiment of the present invention.

FIG. 3 is a block diagram of a reduced latency memory interface for use in a storage apparatus in accordance with a third embodiment of the present invention.

FIG. 4 illustrates a method for using a reduced latency memory interface in a storage apparatus in accordance with a fourth embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

While the invention has been described in conjunction with a specific best mode, it is to be understood that many alternatives, modifications, and variations will be apparent to those skilled in the art in light of the following description. Those skilled in the art would recognize after reviewing this application that the embodiments disclosed, including embodiments directed to methods or processes, can be implemented using various types of integrated circuits, including microprocessors, programmable logic arrays, such as FPGAs, discrete active and passive devices, such as logic gates and logic functions, or any combination of these devices. The use of these devices in the embodiments of the invention shown below would not require undue experimentation or further invention. In addition, for clarity purposes, not all of the routine features of the embodiments described herein are shown or described. It is appreciated that in the development of any such actual implementation, numerous implementation-specific decisions must be made to achieve the developer's specific goals. These specific goals will vary from one implementation to another and from one developer to another. Moreover, it will be appreciated that such a development effort might be complex and time-consuming but would nevertheless be a routine engineering undertaking for those of ordinary skill in the art having the benefit of this disclosure.

Various embodiments of the present invention disclosed herein pertain to achieving reduced memory transaction read latency in a memory device, such as a storage apparatus that employs a memory interface coupled to a memory array. Reduced read latency may be achieved in such a storage apparatus by allocating a buffer from a set of at least one buffer, such as a data register included in a flash device, and by using this allocated buffer to store prefetch data. The number of buffers allocated as prefetch buffers for storing prefetch data may be selected at the time of manufacture, at the time of boot-up or dynamically as needed by the storage apparatus. In this example, this flash device comprises at least a portion of the memory array.

Referring now to FIG. 1, reduced memory read transaction latency may be achieved by using a storage apparatus 10 that includes a memory interface 24 coupled via a memory bus 22-1 to a memory array 26. Memory array 26 includes at least one flash device, such as flash devices 14-1 through 14-n, where n may be any number greater than zero (0). The flash devices may include data registers. For example, flash devices 14-1 and 14-n may respectively include data registers 16-1 through 16-n. The data registers form part of buffer set 28. The term “flash device” is intended to include any form of non-volatile memory that includes a set of non-volatile memory cells, including multi-level memory cells. This flash device permits read and write operations to be performed on these memory cells according to a protocol supported by the flash device. This flash device may be implemented using a flash memory device that complies with the Open NAND Flash Interface Specification, commonly referred to as ONFI Specification. The term “ONFI Specification” is a known device interface standard created by a consortium of technology companies, called the “ONFI Workgroup”. The ONFI Workgroup develops open standards for NAND flash memory devices and for devices that communicate with these NAND flash memory devices. The ONFI Workgroup is headquartered in Hillsboro, Oreg. Using a flash device that complies with the ONFI Specification is not intended to limit the embodiment disclosed. One of ordinary skill in the art having the benefit of this disclosure would readily recognize that other types of flash devices employing different device interface protocols may be used, such as protocols compatible with the standards created through the Non-Volatile Memory Host Controller Interface (“NVMHCI”) working group. Members of the NVMHCI working group include Intel Corporation of Santa Clara, Calif., Dell Inc. of Round Rock, Tex. and Microsoft Corporation of Redmond, Wash.

Memory interface 24 responds to a memory read transaction request 12 received via bus 20 from host 18-1 by performing on memory array 26 a memory operation that corresponds to the transaction requested by host 18-1. For example, if memory interface 24 receives memory read transaction request 12, memory interface 24 performs a read operation on memory array 26 and returns the result of the read operation, such as read data, along bus 20 to host 18-1. Data returned as a result of the read operation is hereinafter called “read data”. Memory interface 24 may also perform an internal read operation if data exists in memory array 26 that meets at least one prefetch selection criterion. This prefetch selection criterion may include a relationship between read data, and a data portion existing in memory array 26, named “stored data”. The type of relationship between read data and stored data is not intended to be limiting in any way. Any type of relationship may be used that increases the likelihood that the stored data selected for prefetching will be later requested by a host coupled to a memory interface 24, such as host 18-1. For example, this prefetch selection criterion may be met if stored data exists that has a memory address that differs from the read data within a selected memory address range. In another example, memory interface 24 determines whether the prefetch selection criterion is met by determining whether any stored data exists in memory array 26 that has a memory address within the same page as a memory address associated with the read data.

The term “host”, such as host 18-1 shown in this embodiment, means any device that has the ability to transmit a memory read transaction request to storage apparatus 10. For example, this may include host 18-1 that can generate a memory read transaction request 12 and that can receive a response resulting from the processing of memory read transaction request 12 by memory interface 24. Memory interface 24 may process memory transactions from more than one requesting device, such as host 18-1 through host 18-n. For example, as shown in FIG. 1, host 18-1 may generate a memory read transaction request 12, while another host 18-n may receive a response, such as prefetch data 30, that is a result of another subsequent memory read transaction request 32 sent by host 18-n.

The term “data portion” is intended to refer to any data size granularity used by storage apparatus 10 to address data stored in memory array 26. For example, data may be addressed in increments that may include a flash page, flash block, a file or the like.

If memory interface 24 determines the prefetch selection criterion is met, it initiates an internal read operation. This internal read operation includes memory interface 24 retrieving data that meets the prefetch selection criterion from memory array 26. Memory interface 24 may perform this internal read operation without host prompting. However, if memory interface 24 determines the prefetch selection criterion has not been met, memory interface 24 does not perform this internal read operation. Data retrieved from memory array 26 that meets the prefetch selection criterion is hereinafter called “prefetch data”.

As part of the internal read operation, memory interface 24 allocates at least one buffer from buffer set 28 to store prefetch data 30. The number of buffers allocated by memory interface 24 as prefetch buffers for storing prefetch data may be selected at the time of manufacture, at the time of initialization or dynamically as needed by the storage apparatus. In one implementation, the number of buffers allocated as prefetch buffers may be initially selected by memory interface 24. Memory interface 24 may then dynamically adjust the number of buffers after initialization, depending on an amount of storage needed for data treated as prefetch data during the processing of memory read transaction requests. For example in FIG. 1, memory interface 24 may allocate data register 16-1 as a storage space suitable for storing prefetch data 30. This action may render non-allocated buffers, such as data registers 16-n in buffer set 28, available to storage apparatus 10 for other memory uses as needed. Buffers allocated for storing prefetch data, such as buffer 16-1 are hereinafter called “prefetch buffers”. Storing prefetch data 30 in a prefetch buffer, such as prefetch buffer 17, renders prefetch data 30 readily available to a requesting host if requested by the host, such as through a subsequent memory read transaction request.

Allocation of a prefetch buffer in the above manner is not to be taken as a limitation in any way. One of ordinary skill in the art having the benefit of this disclosure may use other techniques or procedures to allocate a buffer as a prefetch buffer for storing prefetch data. For example, memory array 26 may further include at least one HDD, which is not shown to avoid overcomplicating this disclosure. During initialization, memory interface 24 operating under program control may by default prefetch data from random areas in this HDD. This configuration option may be preferable when low latency performance during random access is required, such as in database applications. Another option may include configuring memory interface 24 through program code to provide an option in the form of a vendor-specific interface command to allow a host, such as host 18-1, to select the prefetching method to be used by memory interface 24 during initialization. For instance, if storage apparatus 10 will be used to store large files that tend to be stored sequentially, such as video files, memory interface 24 may be configured to prefetch sequential data rather than random data. To reduce initialization time of storage apparatus 10, these prefetching options during initialization may be disabled. In another possible configuration, storage apparatus 10 may support a host-controlled non-volatile cache command set. This allows a host, such as host 18-1, to keep selected data in low latency non-volatile memory that may form a portion of memory array 26, such as flash devices 14-1 through 14-n, rendering this selected data to be available as prefetch data during initialization. During initialization, memory interface 24 operating under the program control detects this selected data and automatically prefetches this data by storing the data in a prefetch buffer.

If memory interface 24 receives memory read transaction request 32, which may be subsequent to memory read transaction request 12, and the data requested by memory read transaction request 32 is currently available in prefetch buffer 17 as prefetch data 30, memory interface 24 will initiate a forwarding transaction, reducing the read latency of the memory read transaction resulting from memory read transaction request 32. This forwarding transaction includes retrieving the prefetch data previously stored in a prefetch buffer, such as prefetch data 30 and prefetch buffer 17, respectively, and forwarding prefetch data 30 to the requesting host, such as host 18-n.

Storage apparatus 10 also includes a data transmission conduit, such as a memory bus 22-1, for coupling memory interface 24 and memory array 26. Memory interface 24 uses memory bus 22-1 as a conduit for transferring data between memory interface 24 and memory array 26. Memory bus 22-1 may include additional memory bus 22-n, where n may be any number greater than zero. Additional memory buses, such as memory bus 22-n, may be implemented to accompany an expansion of memory array 26. Memory bus 22-1 though 22-n are provided as one possible implementation feature and are not intended to be limiting in any way. Bus 20 may be used to couple host 18-1 to storage apparatus 10. Additional buses, such as bus 20-n, may be implemented to enable additional hosts, such as host 18-n, to communicate with memory interface 24.

FIG. 2 illustrates another embodiment of a storage apparatus 34 that supports prefetching transactions and that employs a reduced latency memory interface 48. Storage apparatus 34 includes a memory array 50, which is coupled to memory interface 48 via a memory bus 46. Memory interface 48 is responsive to a memory read transaction request 36 sent via bus 44 from a host 42, and enables storage apparatus 34 to perform reduced latency memory transactions in response to receiving second memory read transaction request 64. In this embodiment, memory interface 48 includes a DMA controller 56 and flash memory controller 60. Buffer set 52 includes data registers 40-1 to 40-n from flash devices 38-1 through 38-n, respectively; at least one buffer from DMA controller 56, such as buffer 58; at least one buffer from flash memory controller 60, such as buffer 62; or any combination of these data registers and buffers. As previously described, a prefetch buffer is a data store that has been allocated to store prefetch data.

The number of buffers allocated as prefetch buffers may be selected at the time of manufacture, during initialization or during the processing of memory operations. In one example, memory interface 48 may initially allocate a selected number of buffers at the time of initialization, and then dynamically adjust the number of buffers allocated as prefetch buffers after initialization, depending on an amount of storage needed for data treated as prefetch data during the processing of memory read transaction requests. This example is not intended to limit the various embodiments disclosed herein. Other embodiments for allocating a selected number of buffers that may be applicable to the example shown in FIG. 2 have been previously disclosed in FIG. 1. In addition, the number of buffers allocated as prefetch buffers may depend on an amount of storage needed for the prefetch data. For instance, memory interface 48 may allocate data register 40-1 and buffer 58 as storage space suitable for storing prefetch data 44. This action may also render non-allocated buffers, such as data registers 40-n and buffer 62 in buffer set 52 available to storage apparatus 34 for other memory uses as needed. Buffers allocated for storing prefetch data, such as buffer 58 and data register 40-1 are hereinafter called “prefetch buffers”.

Memory array 50, host 42-1 through 42-n, flash devices 38-1 through 38-n, bus 46-1 through 46-n, memory read transaction request 36, second memory read transaction request 64 and memory bus 46-1 through 46-n may be implemented to have substantially the same function and structure as memory array 26, host 18-1 through 18-n, flash devices 16-1 though 16-n, bus 20 through 20-n, memory read transaction request 12, second memory read transaction request 32 and memory bus 22-1 through 22-n respectively, disclosed above with respect to FIG. 1.

In this embodiment, buffer set 52 includes buffer 58 and buffer 62. Like data registers 40-1 through 40-n, buffers 58 and 62 form part of buffer set 52 and can be used as prefetch buffers if memory interface 48 allocates these buffers as prefetch buffers. Otherwise, memory interface 24 in FIG. 1 is similar to memory interface 48 in function. For example, when prefetch data 44 is found during an internal read operation, memory interface 48 stores prefetch data 44 into prefetch buffer 35, such as buffer 58 and data register 40-1. Also, if memory interface 48 receives a second memory read transaction request 64, which may be subsequent to the earlier memory read transaction request 36, for data and this data has been previously stored as prefetch data, such as prefetch data 54, and prefetch data 54 is stored in prefetch buffer 35, memory interface 48 initiates a forwarding transaction. This forwarding transaction is similar to the forwarding transaction described earlier with respect to the embodiment corresponding to FIG. 1, above. However, unlike the forwarding transaction described in with respect to FIG. 1 which only shows data registers which may be allocated as prefetch buffers, the forwarding transaction here may include obtaining prefetch data from any of the buffers in buffer set 52 if memory interface 48 allocates those buffers as prefetch buffers.

For example, if the data requested in the second memory read transaction 64 has been stored as prefetch data 44, and prefetch data 44 is stored in buffer 58, the forwarding transaction includes retrieving prefetch data 44 from buffer 58 and forwarding prefetch data 44 to the host that sent the second memory read transaction request. In another example, if the data requested in the second memory read transaction request 64 is stored as prefetch data 44, and prefetch data 44 is stored in data register 40-1, the forwarding transaction includes retrieving prefetch data 44 from data register 40-1 and forwarding prefetch data 44 to the host that sent the second memory read transaction request. Thus, memory interface 48 may perform an internal read operation and forwarding transaction using any of the buffers, such as data registers 40-1 through 40-n, buffer 58, and buffer 62 that memory interface 48 has allocated as prefetch buffers.

FIG. 3 illustrates another embodiment of a storage apparatus 66 that supports prefetching transactions and that employs a reduced latency memory interface 80 in accordance with another embodiment of the present invention. Storage apparatus 66 includes a memory array 82 coupled to a memory interface 80 via a memory bus 98. Memory interface 80 is responsive to a memory read transaction request 68 sent via bus 78 from a host 76-1 by enabling storage apparatus 66 to perform a reduced latency memory transaction in response to receiving a second memory read transaction request 106.

Memory interface 80 includes a DMA controller 88, flash memory controller 92, system bus 96, local memory 100, host interface 102 and local processor 104. Host interface 102 allows at least one device, such as host 76-1, to communicate with devices within storage apparatus 66. These devices may include local processor 104, local memory 100 and DMA controller 88 of memory interface 80. For example, memory read transaction request 68 may be received by memory interface 80 through host interface 102, permitting memory interface 80 to process this request as disclosed herein. System bus 96 is a conduit for transferring data between DMA controller 88, local processor 104, local memory 100 and host interface 102. Local processor 104, host interface 102 and DMA controller 88 may access local memory 100 via system bus 96 as needed. Local memory 100 may be implemented using any form of memory, such as various types of DRAM, non-volatile memory and the like.

Memory array 82 includes at least one flash device, such as flash devices 70-1 through 70-n, where n may be any number greater than zero (0). Each flash device may include flash pages and data registers. For example, flash devices 70-1 and 70-n may each respectively include flash pages 74-0 through 74-n, flash pages 75-0 through 75-n, and data registers 72-1 through 72-n. In this embodiment, data registers form part of buffer set 84. Memory array 82, host 76-1 through 76-n, bus 78 through 78-n, buffer set 84, memory read transaction request 68, second memory read transaction request 106 and memory bus 98-1 through 98-n may be respectively implemented to have substantially the same function and structure as memory array 50, host 42-1 though 42-n, bus 54 through 54-n, buffer set 52, memory read transaction request 36, second memory read transaction request 64 and memory bus 46-1 though 46-n respectively, disclosed above with respect to FIG. 2.

Host 76-1 sends memory read transaction request 68 to storage apparatus 66 for data located in memory array 82. Operating under program control, local processor 104 analyzes and responds to memory read transaction request 68 by generating DMA instructions that will cause DMA controller 88 to read this data from memory array 82 through memory controller 92. If this data is available, memory interface 92 retrieves this data, which is transferred to local memory by DMA controller 88, and eventually transferred to host 76-1 via host interface 102. Data obtained during this memory read transaction request is hereinafter named “read data”.

Local processor 104 also identifies “prefetch data”. Prefetch data is any data currently stored in memory array 82 that meets a prefetch selection criterion. This prefetch selection criterion may include a relationship between the read data and any stored data. For example, this prefetch selection criterion may be met if stored data exists that has a memory address within a selected memory address range from the memory address of the read data. In another example, memory interface 80 determines whether this prefetch selection criterion is met by determining whether any stored data exists in memory array 82 that has a memory address within the same page as a memory address associated with the read data. The type of relationship between read data and stored data is not intended to be limiting in any way. Any type of relationship may be used that increases the likelihood that the stored data selected for prefetching will be later requested by a host coupled to a memory interface 80, such as host 76-1.

Local processor 104 may also perform an additional read operation, called an internal read operation. This internal read operation is a process executed by the local processor 104 to retrieve prefetch data 86 from memory array 82. This internal read operation includes local processor 104 sending an instruction, such as instruction 122, to DMA controller 88 to retrieve prefetch data 86 from memory array 82. For example, if memory array 82 addresses stored data by flash page, instruction 122 contains the address of the specific flash page within a flash device that holds prefetch data 86, such as flash page 74-0 and flash device 70-1, respectively. Instruction 122 also causes DMA controller 88 to retrieve prefetch data 86 from the flash page and flash device identified by the address mentioned above.

When DMA controller 88 receives instruction 122, instruction 122 causes DMA controller 88 to use flash memory controller 92 to cause flash device 70-1 to transfer prefetch data 86. The internal read operation also includes local processor 104 allocating a buffer in buffer set 84 to store prefetch data 86. Allocating a buffer as a prefetch buffer may be performed statically, such as at the time of manufacture, dynamically by local processor 104 or both. In one implementation, the number of buffers allocated as prefetch buffers depends on an amount of storage needed for the prefetch data. For example, in FIG. 3, memory interface 84 may allocate data register 72-1 and buffer 90 as storage space suitable for storing prefetch data 86. This action may also render non-allocated buffers, such as data registers 72-n and buffer 94 in buffer set 84 available to storage apparatus 66 for other memory uses. Buffers allocated for storing prefetch data, such as buffer 90 and data register 72-n, are hereinafter called “prefetch buffers”. If a host, such as host 76-1, issues a second memory read transaction request 106 for data and this data has been previously stored as prefetch data in a prefetch buffer such as prefetch data 86 in prefetch buffer 89, local processor 104 responds by retrieving prefetch data 86 from prefetch buffer 89 and by placing prefetch data 86 into local memory 100. When local processor 104 determines local memory 100 contains prefetch data 86, local processor 104 forwards prefetch data 86 from local memory 100 to host interface 102, which in turn transfers the prefetch data to host 76-1. One of ordinary skill in the art having the benefit of this disclosure would readily recognize that rather than using local memory to store prefetch data, prefetch data may be sent directly to the host or processed in any other manner envisioned by a user.

FIG. 4 discloses a method for reducing latency in memory read transactions in accordance with another embodiment of the present invention. This method may be performed by using a reduced latency memory interface that is coupled to a memory array that includes a flash device and also coupled to a buffer set that includes a buffer. These elements may be provided in a storage apparatus, such as storage apparatus 10, 34, or 66 in FIGS. 1 through 3, respectively.

Under the method shown, a memory read transaction request is received 108. For example, as shown in FIG. 3, a host 76-1 may send a memory read transaction request 68 to storage apparatus 66, which uses host interface 102 to receive memory read transaction request 68. A read operation is then performed 110. This read operation may be performed by memory interface 80 by obtaining read data requested by host 76-1 from memory array 82. For example, in FIG. 3, memory array 24 responds to memory read transaction request 12 and obtains read data from memory array 26, returning read data to host 18-1 if this data is available.

Local processor 104 identifies 112 data currently stored in memory array 82 that meets a prefetch selection criterion. For example, in FIG. 3, if local processor 104 identifies stored data in memory array 82 that are within a selected address range of the address range of the read data.

If prefetch data is identified, an internal read operation is performed 114. For example, in FIG. 3, memory interface 80 causes a flash device in memory array 82 that holds this prefetch data, such as prefetch data 86, to transfer the prefetch data 86 from a flash page storing the prefetch data. Additionally, the internal read operation includes local processor 104 allocating a buffer as a prefetch buffer in buffer set 84 for storing prefetch data 86. The internal read operation further includes storing prefetch data 86 in a prefetch buffer, such as prefetch buffer 89.

If a second memory read transaction request is received for data and that data is stored as prefetch data in a prefetch buffer, a forwarding transaction is performed 120. For example, referring to FIG. 3, if a second memory read transaction request 106 is received for data and that data is stored as prefetch data 86 in prefetch buffer 89, local processor 104 responds to memory read transaction request 106 and retrieves prefetch data 86 from prefetch buffer 89. Local processor 104 forwards prefetch data 86 from prefetch buffer 89 to local memory 100. Local processor 104 forwards prefetch data 86 from local memory 100 to host 76-1. This transfer of prefetch data 86 from buffer set 84 to local memory 100 is not depicted in FIG. 3 to avoid overcomplicating this herein disclosure. In this example, flash devices 14-1 through 14-n function as mass storage devices. However, in another embodiment (not shown), these flash devices may function as a cache to another type of storage device, such as a set of HDDs arranged in a RAID, JBOD or other configuration. This set of HDDs is not shown in the drawings to avoid overcomplicating the herein disclosure, but may be included as part of the memory array, such as memory array 82 in FIG. 4.

While the present invention has been described in particular embodiments, it should be appreciated that the present invention should not be construed as limited by such embodiments. Rather, the present invention should be construed according to the claims below.

Claims

1. A storage apparatus for performing reduced memory read transactions, said storage apparatus responsive to memory transaction requests from at least one host, the storage apparatus comprising: a memory array comprising a plurality of flash memory devices, said plurality of flash memory devices includes a first flash memory device, said first flash memory device including a data register and non-volatile memory cells;a memory interface coupled to said memory array, wherein said memory interface comprises a local processor, a Direct Memory Access (DMA) controller and a flash memory controller, and wherein said DMA controller comprises a first buffer and wherein said flash memory controller comprises a second buffer;a buffer set that includes a plurality of buffers that are suitable for use as prefetch buffers, at least some of said buffer set included in said memory interface;said buffer set comprising said first buffer in said DMA controller in said memory interface, said second buffer in said flash memory controller, and said data register in said first flash memory device such that said first buffer, said second buffer, and said data register are used as said prefetch buffers;wherein said memory interface allocates said prefetch buffers in said buffer set, said prefetch buffers including said first buffer in said DMA controller, said second buffer in said flash memory controller, and said data register so that said buffer set includes said prefetch buffers allocated for storing prefetch data and at least one non-allocated buffer that is available to said storage apparatus for other memory uses;wherein said memory interface is configured to identify stored data that meets a prefetch selection criterion and wherein said stored data are stored in said memory array;wherein, in response to receiving a memory read transaction request, said memory interface performs a read operation wherein said DMA controller reads said stored data from said memory array through said flash memory controller and, if said memory interface identifies said stored data that meets said prefetch selection criterion in said memory array, said memory interface performs an internal read operation that includes allocating said prefetch buffers in said buffer set and storing said stored data in at least one of said prefetch buffers as said prefetch data; andwherein, if said memory interface receives a second memory read transaction request for read data and if said read data is currently available as said prefetch data in at least one of said prefetch buffers, said memory interface responds to said second memory read transaction request by performing a forwarding transaction that includes said local processor retrieving said prefetch data from at least one of said prefetch buffers and said local processor forwarding said prefetch data to said host, reducing read latency of said second memory read transaction request.

2. The storage apparatus of claim 1, wherein said DMA controller reads data from said first flash memory device in said memory array through said flash memory controller.

3. The storage apparatus of claim 1, wherein said memory interface further includes a host interface for receiving said memory read transaction request from said host.

4. The storage apparatus of claim 1, wherein said DMA controller uses said flash memory controller to transfer said prefetch data from said memory array.

5. The storage apparatus of claim 1, wherein said buffer set comprises said first buffer which comprises a first fixed buffer from said DMA controller.

6. The storage apparatus of claim 1, wherein said buffer set includes a second fixed buffer from said flash memory controller.

7. The storage apparatus of claim 1, wherein said first flash memory device contains at least one flash page.

8. The storage apparatus of claim 1, wherein said storage apparatus includes a memory bus for coupling said memory array to said memory interface.

9. The storage apparatus of claim 1, wherein said storage apparatus includes a bus for receiving said memory read transaction request from said host and sending data from the storage apparatus to said host.

10. The storage apparatus of claim 1, wherein said storage apparatus includes a system bus for receiving and sending data between said local processor, said DMA controller, a local memory, and a host interface.

11. The storage apparatus of claim 1 wherein said memory interface includes said DMA controller, said local processor, a host interface and a local memory coupled together through a system bus; and said memory interface further includes said flash memory controller coupled to said DMA controller and to said memory array through a memory bus.

12. The storage apparatus of claim 1, wherein said buffer set includes another data register in another flash memory device in said memory array.

13. A method for reducing latency in memory read transactions, the method comprising: using a low latency memory interface coupled to a memory array comprising a plurality of flash memory devices, said plurality of flash memory devices includes a first flash memory device, said first flash memory device includes a data register and non-volatile memory cells;wherein said memory interface comprises a buffer set that includes a plurality of buffers that are suitable for use as prefetch buffers, wherein said memory interface comprises a local processor, a Direct Memory Access (DMA) controller and a flash memory controller, and wherein said DMA controller comprises a first buffer and wherein said flash memory controller comprises a second buffer;wherein said buffer set comprises said first buffer, said second buffer, and said data register such that said first buffer, said second buffer, and said data register are used as said prefetch buffers;allocating, by said memory interface, said prefetch buffers in said buffer set, said prefetch buffers including said first buffer in said DMA controller, said second buffer in said flash memory controller, and said data register so that said buffer set includes said prefetch buffers allocated for storing prefetch data and at least one non-allocated buffer that is available to said memory interface for other memory uses; andwherein said memory interface responds to memory read transaction requests from a host, the method further comprising: receiving, by said memory interface, a memory read transaction request;performing a read operation in response to said memory read transaction request wherein said DMA controller reads stored data from said memory array through said flash memory controller;identifying, by said memory interface, said stored data that meets a prefetch selection criterion, wherein said stored data are stored in said memory array;if said memory interface identifies said stored data that meets said prefetch selection criterion in said memory array, then performing an internal read operation that includes allocating said prefetch buffers in said buffer set and storing said stored data in at least one of said prefetch buffers as said prefetch data; andif said memory interface receives a second memory read transaction request for read data and if said read data is currently available as said prefetch data in at least one of said prefetch buffers, then performing a forwarding transaction that includes said local processor retrieving said prefetch data from at least one of said prefetch buffers and said local processor forwarding said prefetch data to said host.

14. The method of claim 13 wherein said prefetching selection criterion is met if said prefetch data is associated with a memory address within a same page as a memory address associated with said read data.

15. The method of claim 13 wherein said prefetch selection criterion is met if said prefetch data is associated with a memory address within a same flash memory device as a memory address associated with data read during a response to said memory read transaction request.

16. The method of claim 13 wherein said prefetch selection criterion is met if a memory address of said prefetch data differs within a selected memory address range from a memory address of data read during a response to said memory read transaction request.

17. The method of claim 13 wherein a response to said second memory read transaction request includes transferring said prefetch data from at least one of said prefetch buffers to a local memory.

18. The method of claim 17 wherein a response to said second memory read transaction request includes forwarding said prefetch data from said local memory to said host.

19. An article of manufacture, comprising: a non-transitory computer-readable medium having stored thereon instructions operable to permit an apparatus to perform a method comprising: receiving, by a memory interface, a memory read transaction request from a host;wherein said memory interface is coupled to a memory array comprising a plurality of flash memory devices, said plurality of flash memory devices includes a first flash memory device, wherein said first flash memory device comprises a data register and non-volatile memory cells;wherein said memory interface comprises a buffer set that includes a plurality of buffers that are suitable for use as prefetch buffers;wherein said memory interface comprises a local processor, a Direct Memory Access (DMA) controller and a flash memory controller, and wherein said DMA controller comprises a first buffer and wherein said flash memory controller comprises a second buffer;wherein said buffer set comprises said first buffer, said second buffer, and said data register such that said first buffer, said second buffer, and said data register are used as said prefetch buffers;allocating, by said memory interface, said prefetch buffers in said buffer set, said prefetch buffers including said first buffer in said DMA controller, said second buffer in said flash memory controller, and said data register so that said buffer set includes said prefetch buffers allocated for storing prefetch data and at least one non-allocated buffer that is available to said memory interface for other memory uses;performing a read operation in response to said memory read transaction request wherein said DMA controller reads stored data from said memory array through said flash memory controller;identifying, by said memory interface, said stored data that meets a prefetch selection criterion, wherein said stored data are stored in said memory array;performing an internal read operation that includes allocating said prefetch buffers in said buffer set and storing said stored data in at least one of said prefetch buffers as said prefetch data if said memory interface identifies said stored data that meets said prefetch selection criterion in said memory array; andperforming a forwarding transaction that includes said local processor retrieving said prefetch data from at least one of said prefetch buffers and said local processor forwarding said prefetch data to said host if said memory interface receives a second memory read transaction request for read data and if said read data is currently available as said prefetch data in at least one of said prefetch buffers.

20. The article of manufacture of claim 19, wherein said DMA controller reads data from said first flash memory device in said memory array through said flash memory controller.

21. A storage apparatus for performing reduced latency memory read transactions, the storage apparatus comprising: a memory array including a plurality of non-volatile memory devices;a memory interface coupled by a memory bus to the memory array and responsive to memory read transaction requests;the memory interface coupled by a bus to a host during a transmission of memory transaction requests from the host to the memory interface;the memory interface including a Direct Memory Access (DMA) controller comprising a first buffer and a flash memory controller comprising a second buffer; andwherein the memory interface allocates a plurality of prefetch buffers in a buffer set, the plurality of prefetch buffers including the first buffer in the DMA controller and the second buffer in the flash memory controller so that the buffer set includes the plurality of prefetch buffers allocated for storing prefetch data and one or more non-allocated buffers available to the storage apparatus for other memory uses;the memory interface configured to perform a read operation on the memory array in response to a memory read transaction request and configured to perform an internal read operation if the memory interface identifies stored data, in the memory array, meeting a prefetch selection criterion;the internal read operation comprising the memory interface allocating a data store in the buffer set and storing the stored data in the data store as the prefetch data available to a second memory read transaction.

22. A method of performing reduced latency memory read transactions in a storage apparatus, the method comprising: receiving, by a memory interface, a memory read transaction request from a host and performing, by the memory interface, a read operation on a memory array; andperforming, by the memory interface, an internal read operation if stored data, in the memory array, meets a prefetch selection criterion;wherein performing the internal read operation comprises: allocating a data store in a buffer set and storing the stored data in the data store as a prefetch data;the memory array including a plurality of non-volatile memory devices;the memory interface coupled by a bus to the host during a transmission of memory transaction requests from the host to the memory interface;the memory interface including a Direct Memory Access (DMA) controller comprising a first buffer and a flash memory controller comprising a second buffer;wherein the memory interface, coupled by a memory bus to the memory array, allocates a plurality of prefetch buffers in the buffer set, wherein the plurality of prefetch buffers includes the first buffer in the DMA controller and the second buffer in the flash memory controller so that the buffer set includes the plurality of prefetch buffers allocated for storing the prefetch data available to a second memory read transaction and one or more non-allocated buffers available to the storage apparatus for other memory uses.

23. An article of manufacture, comprising: a non-transitory computer-readable medium having stored thereon instructions operable to permit a storage apparatus to perform a method comprising: receiving, by a memory interface, a memory read transaction request from a host and performing, by the memory interface, a read operation on a memory array; andperforming, by the memory interface, an internal read operation if stored data, in the memory array, meets a prefetch selection criterion;wherein performing the internal read operation comprises: allocating a data store in a buffer set and storing the stored data in the data store as a prefetch data;the memory array including a plurality of non-volatile memory devices;the memory interface coupled by a bus to the host during a transmission of memory transaction requests from the host to the memory interface;the memory interface including a Direct Memory Access (DMA) controller comprising a first buffer and a flash memory controller comprising a second buffer;wherein the memory interface, coupled by a memory bus to the memory array, allocates a plurality of prefetch buffers in the buffer set, wherein the plurality of prefetch buffers includes the first buffer in the DMA controller and the second buffer in the flash memory controller so that the buffer set includes the plurality of prefetch buffers allocated for storing the prefetch data available to a second memory read transaction and one or more non-allocated buffers available to the storage apparatus for other memory uses.