Data writing method and device

Abstract

The present invention proposes a data writing method and device. The data writing method includes: receiving a write operation request sent by an application; storing the write operation request into a first cache; when the type of the write operation request is write immediate When returning or writing and synchronously operating, or when the first cache has met the cache switching condition, the cache for storing the received write operation request is switched from the first cache to the second cache; for the first cache The write operation requested to perform a write operation. The invention can give full play to the parallelism of batch reading and writing and disk writing operations, and can maximize throughput without affecting the semantics of application writing operations.

Description

technical field [0001] The invention relates to the technical field of the Internet, in particular to a data writing method and device. Background technique [0002] In a high-throughput storage system, the pressure of reading and writing often reaches tens of thousands of requests per second, and the amount of requested data varies in size. Especially in structured storage engines, write requests are often only for modifying or updating a few effective short attribute fields, and require immediate read visibility after updating to ensure session transactionality. [0003] At the same time, more and more systems adopt a log-based system design (Write Append Logging; hereinafter referred to as: WAL) design to convert all writes into sequential writes as much as possible, so as to reduce the movement of the magnetic head and persist data. Since the WAL design needs to transfer data to the disk in real time, which requires calling the synchronization (sync) interface of the op...

AI Technical Summary

Problems solved by technology

In the current Serial Advanced Technology Attachment (Serial Advanced Technology Attachment; hereinafter referred to as: SATA) disk environment, since the minimum unit of each write is a sector, if the amount of data is too small, the excess space in the sector will be will be wasted

In the driver layer or operating system layer, the write cache is usually a page size (4K or 8K). Therefore, if a write operation is called for each update or write of a small amount of data, the delay will be increased, and it is difficult to improve the throughput. quantity

Images