56results about How to "Reduce the number of write operations" patented technology

The invention discloses a phase change memory write operation method. The method comprises the following steps: S1, preprocessing a data block to be written back to a PCM storage unit to acquire a preprocessed data block; S2, reading an initial data block corresponding to the data block to be written from the PCM storage unit; and S3, calculating a Hamming distance between the data block to be written or the preprocessed data block and the initial data block, selecting the data block corresponding to the minimum value of the Hamming distance to the initial data block, and writing the data block back to the initial data block, and writing a preprocessing mark back to the storage unit. The method reduces the times of write operation on the PCM storage unit, so the service life of the PCM is prolonged.

The invention discloses a high frequency write operation method comprising the following steps: counting write operation frequency; combining the high frequency write operations in one or more write operations if the write operation is determined to be high frequency write operation; writing the one or more write operation data in a storage server. The technical scheme can be employed to combine high frequency write operations into low frequency write operations, thus reducing write operation times, and effectively saving network bandwidth and storage server used by the high frequency write operations.

The invention discloses a data distribution method of a decentralized distributed heterogeneous storage system, which comprises the following steps: 1. Classifying data objects; 2. Classifying storage devices; 3. Dividing stored data into different "placement groups" Each type of storage device corresponds to a type of "placement group cluster"; 4. Calculate the proportion of each data object to be stored in a different type of "placement group cluster"; 5. Use the hash algorithm to determine Which "placement group" in the "placement group cluster" the data object to be stored belongs to; 6. Use the data distribution algorithm of the storage system to store the data objects in each "placement group" to multiple corresponding storage devices Medium; 7. During the operation of the system, the migration threshold is calculated according to the access characteristics of the data objects, and the data objects are dynamically migrated. The invention has the advantages of maintaining the performance, load balance and expansibility of the storage system, and reducing the number of write operations to the solid-state hard disk.

The embodiment of the invention discloses a distributed system data writing method and device which are used for reducing the burden of data writing operation and the network overhead of verification so as to improve the overall performance of a distributed system. The method comprises the steps that after at least two data writing requests including a first data writing request and a second data writing request are received, a temporary partition is determined according to first to-be-written data and a first attribution partition carried by the first data writing request and second to-be-written data and a second attribution partition carried by the second data writing request; and the method also includes writing the first to-be-written data and the second to-be-written data into temporary partitions, and marking the temporary partition and the first attribution partition corresponding to the first to-be-written data written into the temporary partitions and the temporary partition and the second attribution partition corresponding to the second to-be-written data written into the temporary partitions; further, according to the temporary marks, writing the first to-be-written data back to the first attribution partition, and writing the second to-be-written data back to the second attribution partition.

The invention discloses an optimization method for improving the writing performance of 3D flash memory chips in a solid state disk. The invention takes a hybrid solid state disk with 3D-flash and MLC as storage media as a reference, and at the same time assumes that the size of a single data page of 3D-flash is 16K And the size of a single MLC data page is 4K, this solution can make full use of the feature that the size of the write request does not reach the size of one page, thereby reducing the number of write operations of the flash memory and effectively extending the service life of the flash memory.

The invention discloses a PCM (pulse code modulation) write operation method. The method comprises the following steps: S1, preprocessing a data block to be written back to a PCM storage unit to acquire a preprocessed data block; S2, reading an initial data block corresponding to the data block to be written from the PCM storage unit; and S3, calculating a Hamming distance between the data block to be written or the preprocessed data block and the initial data block, selecting the data block corresponding to the minimum value of the Hamming distance to the initial data block, and writing the data block back to the initial data block, and writing a preprocessing mark back to the storage unit. The method reduces the times of write operation on the PCM storage unit, so the service life of the PCM is prolonged.

The invention discloses a method for separating garbage collection data and user data in a solid state disk, comprising: allocating an active block in the solid state disk as a special block for storing garbage collection data, and setting the thermal attribute value of the special block to 0; Allocate N active blocks in the solid state disk as user blocks for storing user data, and set the thermal attribute values ​​of N user blocks in turn from 1 to N; when a write request arrives, if the write request is generated by garbage collection, the dedicated block Process garbage collection data; if the write request is generated by a user write request, determine the thermal attribute value corresponding to the user data according to the degree of hotness and coldness of the user data, and process the user data by the user block with the same thermal attribute value. The invention realizes the separation of garbage collection data and user data, effectively reduces the number of write operations caused by garbage collection, and further effectively improves the service life and performance of the NAND solid state disk.

The invention discloses a method and system for ensuring data crash consistency. The method includes the steps that the writing frequency of a first cache line corresponding to a write request is determined; when the writing frequency of the first cache line is larger than or equal to a threshold, data corresponding to the write request is written in a second cache line of a nonvolatile cache memory; when the writing frequency of the first cache line is smaller than the threshold, the data in the first cache line is migrated into a nonvolatile memory, and after the data of the first cache line is migrated, data corresponding to the write request is written in the first cache line; when a memory transaction where the write request belongs is completed, the data in the first cache line is marked as invalid data and the data in the second cache line is marked as new valid data if the data of the first cache line is not migrated; if the data of the first cache line is migrated, the data in the first cache line is marked as new valid data. By means of the method and system, the number of times of writing operation on the relatively slow nonvolatile memory is reduced, and the performance of a storage system is improved.