Distributed storage system and configuration method of distributed storage system

Abstract

The invention discloses a distributed storage system and a configuration method of the distributed storage system. The system comprises a first storage pool and a second storage pool, wherein the first storage pool and the second storage pool are respectively composed of a plurality of arrangement groups, and the arrangement groups are basic composition units of the storage pools. Each arrangementgroup comprises a main arrangement group and a slave arrangement group corresponding to the main arrangement group, wherein all the main homing groups of the first storage pool are distributed on object storage equipment of a first data center, the slave homing groups of the first storage pool are distributed on object storage equipment of the first data center and object storage equipment of a second data center, and the second data center and the first data center are backups for each other; all the master arrangement groups of the second storage pool are distributed on the object storage devices of the second data center, and the slave arrangement groups of the second storage pool are distributed on the object storage devices of the first data center and the second data center. The technical problem that an existing distributed storage system lacks an intra-city double-data-center disaster tolerance scene scheme is solved.

Description

technical field [0001] The present application relates to the field of performance optimization of a distributed storage system, and in particular, relates to a distributed storage system and a configuration method of the distributed storage system. Background technique [0002] Ceph is currently the most widely used open source distributed storage system and has been supported by many manufacturers. Many commercial distributed storage systems are deeply customized based on Ceph. Moreover, Ceph has become the "standard" cloud native storage system of the OpenStack cloud platform. [0003] Ceph adopts the CRUSH algorithm, with balanced data distribution and high parallelism. There is no central node in this architecture. The client calculates the location where it writes data through a device mapping relationship, so that the client can directly communicate with the storage node, thereby avoiding the performance bottleneck of the central node. [0004] The following uses a...

AI Technical Summary

Problems solved by technology

[0009] image 3 It is a schematic diagram of a distributed storage system with dual data centers in the same city according to an embodiment of the present application, such as image 3 As shown, but when distributed storage is used in a disaster recovery solution with two data centers, due to the physical distance between data centers, there is usually a certain network delay, which is usually within 10ms round trip, which will be extremely Greatly affects the performance of high-frequency small-block IO. If it is 10ms, it will reduce the performance by more than half, and even affect the normal operation of the business when it is larger.

[0010] Second, the performance of read operations

[0012] Third, if Figure 5 As shown, when a failover occurs on the application side of the Openstack cloud platform, the main pg of the distributed storage needs to be able to follow the application side to perform the corresponding pg switch master. Since the redistribution of pg may cause data reconstruction, it is not expected that this reconstruction operation will affect to application business performance

[0013] Aiming at the problem that the current distributed storage system lacks disaster recovery scenarios with dual data centers in the same city, no effective solution has been proposed yet

Images