摘要:
A system (100, 300) like a video on demand server has a storage medium (106) with a number of files (200) from which data can be fetched or in which data can be stored. A file contains a number of clusters (202, 204, 206) and each cluster contains a number of storage sectors (208) that are physically successively organized on the storage medium. When a new stream is requested, either for reading or writing, a block size (210) is determined for the new stream. The block size is large enough to support the rate of the stream and is chosen such that a cluster can be read with an integer number of blocks by evenly partitioning the size of the cluster into the size of the blocks reading the cluster.
摘要:
The invention relates to a method of writing data to and reading data from a memory (102) by a host. By limiting processing time, time restrictions regarding the data to be processed can be met. However, this is at the expense of data integrity. By setting time limits for a group (300) of multiple processing assignments and dynamic allocation of time, reserved for error recovery and retries, data integrity can be improved, while time limits can still be met. Furthermore, by attaching priorities to the assignments, multiple types of data can be handled in one group of assignments. For example, assignments with real-time requirements can be put in a group with assignments with best effort requirements. In this way, more flexible planning of processing assignments is possible. The invention also relates to a system for writing data to and reading data from a memory.
摘要:
A method for breathing of scheduling algorithms for a storage device (110). The method including: (a) computing a worst-case duration of a breathing cycle (P) for the storage device (110); (b) starting a breathing cycle; (c) determining if one of the following becomes true before the end of P: (i) a number of real-time requests is at least a predetermined threshold based on a number of data streams and performance parameters of the storage device; and (ii) a number of pending requests for any single stream becomes more than one; (d) if at least one of (i) and (ii) remain true during the duration of P, starting a subsequent breathing cycle after completion of the breathing cycle; and (e) if both of (i) and (ii) are not true during the duration of P, waiting P time units from the start of the breathing cycle before starting the subsequent breathing cycle.
摘要:
A computer system includes a system memory (SM) and a non-volatile memory (NVM). The computer system is arranged to copy at least a part of the system state, stored in the system memory during a clean boot-up procedure, into the non-volatile memory (209, 329). During subsequent start-ups of the computer system, the system state is copied from the non-volatile memory into the system memory (205, 305), resulting in a significantly faster system start-up. In case the configuration of the computer system has changed, the complete boot-up procedure is executed again and the new system state is copied into the non-volatile memory, overwriting the previously stored system state.
摘要:
The present invention relates to a write controller for a memory with a plurality of non-volatile storage cells, a read controller for a memory with a plurality of nonvolatile storage cells, to a combined write/read controller, to a solid state device comprising a memory with a plurality of non-volatile storage cells, a programmer device for writing a binary code to a non-volatile memory, to a method for writing data comprising at least one input bit to a memory having non-volatile storage cells, and to a method for controlling the integrity of data comprising at least one input bit stored in non-volatile storage cells of a memory. The basic concept of the present invention is to extend information stored in a non-volatile memory by at least one checking bit. The checking bit is allocated to one code bit, or to each of a plurality of code bits.
摘要:
There is provided a method of increasing boot-up speed in a computer system (10). The system (10) includes computing devices (20) for processing data and a data store (60) coupled thereto for providing data to and receiving data from the devices (20). The store (60) is operable to write and/or read data in several regions (T1, T2, T3) of a data medium (200). Access between the regions is subject to associated jump delays (SK1/2, SK2/3; SK1/3, SL3/2). The store (60) includes a cache (320) for temporarily storing data read from and/or for writing data to the medium (200). On initial boot-up of the system (20), a log is made of a sequence in which the regions (T1, T2, T3) are accessed. Moreover, on subsequent boot-up of the system (10), the log is used to store data read from the medium (200) temporarily in the cache (310, 320) so as to provide for a more temporally efficient sequence of accessing the regions (T1, T2, T3) for speeding up said subsequent boot-up.
摘要:
A universal memory device is presented that provides adaptability to existing hardware and software environments. The memory can “mimic” existing memory technology combining the advantages of integration all memory capacity in to one single technology and still providing the implicit protections and access characteristics known from the different existing memory technologies. The memory device comprises a memory having a plurality of low-latency, rewritable, non-volatile memory cells forming at least one memory section, a profile storage unit connected with said memory and comprising access information allocated to at least one set of request information elements (r: quest profile), such that said access information indicates whether a request for access to said memory (access request), said access request having said request profile, is to be allowed or rejected, and an access control unit communicating with said profile storage unit and said memory, and adapted to allow or reject an incoming access request in dependence on the access information allocated to the request profile of the access request.
摘要:
A universal memory device is presented that provides adaptability to existing hardware and software environments. The memory can “mimic” existing memory technology combining the advantages of integrating all memory capacity into one single technology and still providing the implicit protections and access characteristics known from the different existing memory technologies. The memory device comprises a memory having low-latency, rewritable, non-volatile memory cells, a profile storage unit connected with the memory and comprising access information allocated to a set of request information elements (request profile), such that the access information indicates whether an access request to said memory, the access request having the request profile, is to be allowed or rejected, and an access control unit communicating with the profile storage unit and the memory, and adapted to allow or reject an incoming access request in dependence on the access information allocated to the request profile of the access request.
摘要:
The present invention relates to a write controller (10) for a memory with a plurality of non-volatile storage cells, a read controller for a memory with a plurality of nonvolatile storage cells, to a combined write/read controller, to a solid state device comprising a memory with a plurality of non-volatile storage cells, a programmer device for writing a binary code to a non-volatile memory, to a method for writing data comprising at least one input bit to a memory having non-volatile storage cells, and to a method for controlling the integrity of data comprising at least one input bit stored in non-volatile storage cells of a memory. The invention provides a reliable detection of changes that have occurred to the content of a non-volatile memory. The basic concept of the present invention is to extend information stored in a non-volatile memory by at least one checking bit. The checking bit is allocated to one code bit, or to each of a plurality of code bits. The allocation is preferably reflected in an allocation of memory cells storing the input and code bits.
摘要:
The present invention provides a method and system for reducing fragmentation (both file fragmentation and free-space fragmentation) in digital storage devices (10). In such a device in which a plurality of files is available for deletion the invention proposes selectively deleting the file that maximises defragmentation. The selection may optimise file and free-space defragmentation.