CN106681649A

CN106681649A - Distributed storage metadata accelerating method

Info

Publication number: CN106681649A
Application number: CN201510745386.3A
Authority: CN
Inventors: 杨造文
Original assignee: Hunan Bailimu Technology Co Ltd
Current assignee: Hunan Bailimu Technology Co Ltd
Priority date: 2015-11-06
Filing date: 2015-11-06
Publication date: 2017-05-17

Abstract

The invention relates to a distributed storage metadata accelerating method. According to the method, solid-state discs are configured through a computer storage server, mirror images of the two solid-state discs become a virtual secondary caching disc, and the secondary caching disc is arranged between cache for business reading and writing and a memory of a physical hard disc, is used for processing storage of metadata information of business access data, and is synchronized to storage nodes of other computers in a distributed manner. In the method, capability of high random reading and writing performance of the solid-state discs and the characteristic that data are not lost after power failure are utilized, and rapid reading, writing and synchronization of metadata can be realized effectively to improve reading and writing performance of a distributed storage system.

Description

A kind of distributed storage metadata accelerated method

Technical field

Patent of the present invention is related to Computer Storage field.

Background technology

As informationalized huge explosion and data exponentially level increase, the storage system of processing data is adopted During with traditional single control or dual controller framework, traffic handing capacity cannot much complete mass data Process demand, and distributed memory system is all many times to sacrifice certain performance to guarantee data Strong consistency；The appearance of solid-state disk solves the problems, such as well performance bottleneck, but the price of solid-state disk It is high so that compartment system storage system is too high using the solution cost of all solid state disk.

Therefore be highly desirable to design a kind of mechanism of L2 cache, be applied to distributed memory system with Lift the read-write of service metadata and be synchronized to other distributed storage nodes, to lift Business Processing energy Power.

Patent of invention content

Patent of the present invention is related to a kind of distributed storage metadata accelerated method, and the method passes through computer Storage server configures solid-state disk, and two pieces of solid-state disk mirrorings are a virtual L2 cache disk, and this two Level disc cache is arranged between the caching of business read-write and physical hard disk storage, for processing business access The storage of the metadata information of data and distributed synchronization are to other Computer Storage nodes, the method profit With the ability of the high random read-write performance of solid-state disk, and the characteristics of data are not lost after power down, can have Fast reading and writing and the synchronization of realizing metadata of effect, to lift the business property read and write of distributed memory system Energy.

Description of the drawings

Fig. 1 is a kind of distributed storage metadata accelerated method structural representation of patent of the present invention.

Specific embodiment

In order that the purpose of patent of the present invention, technical scheme and advantage become more apparent, below in conjunction with Drawings and Examples, are further elaborated to patent of the present invention.It should be appreciated that this place is retouched The specific embodiment stated only to explain patent of the present invention, is not intended to limit the present invention patent.

Fig. 1 is referred to, Fig. 1 is a kind of distributed storage metadata accelerated method structure of patent of the present invention Schematic diagram.

A kind of distributed storage metadata accelerated method, it is characterised in that the hardware of methods described is constituted For Computer Storage server (14a), (14b) and (14c) more than three, and solid-state disk (12a), (12b) with (12c), and physical hard disk (15), the logic components of methods described are constituted solid-state disk L2 cache (13a), (13b) and (13c) after mirror image, the business datum of methods described is constituted For metadata (10a) and (10b), business datum (11a) and (11b).

A kind of distributed storage metadata accelerated method, it is characterised in that the business of methods described writes stream Journey is service metadata (10a) and business datum (11a) when distributed memory system is written to, industry Business metadata (10a) can first write the virtual L2 cache (13a) of solid-state disk, then same by (13a) Step write is all (13b) and (13c) of L2 cache, while business datum (11a) write is calculated In machine memory node (14a), treat that data write is returned to after the completion of two above step to be completed, while Backstage is index according to the information of service metadata (10a), and finishing service data (11a) are synchronously written To in Computer Storage node (14b) and (14c).The service metadata (10b) and business number Business processing flow according to (11b) is identical with upper, difference be processing business metadata (10b) and The main L2 cache of business datum (11b) is (13b) and Computer Storage node is (14b).

A kind of distributed storage metadata accelerated method, it is characterised in that the business of methods described reads stream Journey is first to retrieve the virtual L2 cache (13a) of solid-state disk, such as retrieves service metadata (10a), The business datum (11a) in Computer Storage node (14a) is then directly read with finishing service read request, To other Nodes Retrieval service metadatas (10a) if not retrieving, other are processed reads flow process ibid.

A kind of distributed storage metadata accelerated method, it is characterised in that methods described is practically applicable to simultaneously Distributed storage node is stored as the scene of two parts of copy datas and three parts of copy datas.

The preferred embodiment of patent of the present invention is the foregoing is only, it is of the invention not special to limit Profit, all any modification, equivalent and improvement made within the spirit and principle of patent of the present invention Deng should be included within the protection domain of patent of the present invention.

Claims

1. a kind of distributed storage metadata accelerated method, it is characterised in that the hardware structure of methods described Become Computer Storage server (14a), (14b) and (14c) more than three, and solid-state disk (12a), (12b) with (12c), and physical hard disk (15), the logic components of methods described are constituted solid-state disk L2 cache (13a), (13b) and (13c) after mirror image, the business datum of methods described is constituted For metadata (10a) and (10b), business datum (11a) and (11b).

2. a kind of distributed storage metadata accelerated method according to claim 1, its feature exists In it is that service metadata (10a) and business datum (11a) are writing that the business of methods described writes flow process During to distributed memory system, service metadata (10a) can first write the virtual L2 cache of solid-state disk (13a) (13b) and (13c) for being all L2 cache, then by (13a) is synchronously written, while industry In business data (11a) write Computer Storage node (14a), treat just to be returned after the completion of two above step Return data write to complete, while backstage is index according to the information of service metadata (10a), complete industry Business data (11a) is synchronously written in Computer Storage node (14b) and (14c).The business The business processing flow of metadata (10b) and business datum (11b) is identical with upper, and difference is place The main L2 cache of reason service metadata (10b) and business datum (11b) is (13b) and computer Memory node is (14b).

3. a kind of distributed storage metadata accelerated method according to claim 1, its feature exists In it is first to retrieve the virtual L2 cache (13a) of solid-state disk that the business of methods described reads flow process, is such as examined Rope then directly reads the business datum in Computer Storage node (14a) to service metadata (10a) (11a) with finishing service read request, to other Nodes Retrieval service metadatas (10a) if not retrieving, Other are processed reads flow process ibid.

4. a kind of distributed storage metadata accelerated method according to claim 1, its feature exists In simultaneously methods described is practically applicable to distributed storage node and is stored as two parts of copy datas and three parts of copies The scene of data.