Method and device for writing in data in distributed system

Abstract

The objective of the invention is to provide a method and a device for writing in data in a distributed system. The method comprises steps of requesting for distribution of multiple slave node devices for writing in data for a master node device at a client node device end; then, simultaneously writing the current data in each currently distributed slave node device; acquiring a write-in result fed back by each currently distributed slave node device, and judging whether the current data has been totally written into each slave node device or not based on the write-in result; and if not, circularly executing above processes until the current data has been totally written into each currently distributed slave node device. Thus, communication frequency between devices when the data is written is effectively reduced. Meanwhile, by simultaneously writing data blocks in each currently distributed slave node device in an asynchronous request manner, the delay in writing in is greatly reduced, and the throughput efficiency of the distributed system is increased.

Description

technical field [0001] The present application relates to the computer field, and in particular to a technology for writing data in a distributed system. Background technique [0002] At present, the distributed storage system uses a large number of PC (personal computer) clusters to store massive data. In order to ensure data security, three copies of storage are generally carried out on different PCs. The loss of a copy will not result in the loss of user data. With the development of Internet technology, massive file storage has higher and higher requirements for storage technology and systems. For example, when storing logs, data must be persisted when writing data, and the read and write delay of logs restricts the throughput of the entire system. Therefore, in general Logging systems all require extremely low latency, not only requiring low average latency, but also low latency glitch rate. [0003] In the prior art, a common way of writing data in a distributed stor...

AI Technical Summary

Problems solved by technology

[0003] In the existing technology, the common way of writing data in distributed storage systems is usually chain writing. A successful write requires at least six serial communications between machines, and the delay is relatively high.

Images