CN109739696A

CN109739696A - Double-control storage array solid state disk cache acceleration method

Info

Publication number: CN109739696A
Application number: CN201811522575.4A
Authority: CN
Inventors: 田鹏; 马建鹏; 杨扬; 万琨
Original assignee: Beijing Institute of Computer Technology and Applications
Current assignee: Beijing Institute of Computer Technology and Applications
Priority date: 2018-12-13
Filing date: 2018-12-13
Publication date: 2019-05-10
Anticipated expiration: 2038-12-13
Also published as: CN109739696B

Abstract

The invention relates to a dual-control storage array solid-state hard disk cache acceleration method, and relates to the technical field of computer storage. The invention provides a cache acceleration method for dual storage controllers, which realizes that after one storage controller fails, the data storage operation is taken over by another controller, that is, the cache system can still continue to work in the other controller, and data storage is guaranteed. Service is not interrupted. The cache index trees in the two storage controllers are synchronized in real time and are in a mirrored state. When one storage controller fails, the other controller can seamlessly take over storage services on that controller at the point of failure. The invention not only ensures the high-efficiency performance of the SSD cache storage, but also ensures the high reliability of the SSD cache storage.

Description

A kind of dual control storage array solid state hard disk caching accelerated method

Technical field

The present invention relates to computer memory technical fields, and in particular to a kind of dual control storage array solid state hard disk caching acceleration Method.

Background technique

Traditional mechanical hard disk (HDD) has large capacity advantage, but performance is relatively relatively low, especially random I/O performance, The performance bottleneck for often becoming system, show under virtualized environment and becomes apparent, because virtualizing scene, to aggravate I/O random Change.Compared to HDD, solid state hard disk (SSD) advantage with high performance, especially in terms of random I/O, advantage clearly, still The hardware cost of SSD is relatively high.

Industry has done some optimizations in terms of combining the high-performance of large capacity and SSD of HDD at present, basic ideas be using Caching of the SSD as HDD.Dsc data is cached by SSD, when storage control reads and writes data, once hit caching, just directly It is read and write from caching.In SSD cold data caching can according to it is certain strategy under brush in HDD.It is utilized HDD's to reach Large capacity advantage, and promote the purpose of HDD performance.

In the storage system of double storage controllers, since the metadata being buffered in controller memory will not periodic refreshing To SSD, once master controller breaks down, storage service is switched to from controller, cache metadata do not update to from Controller can not reorganize caching from controller, to cause loss of data.Therefore, the caching for realization dual controller adds Speed will guarantee that the metadata of caching is lower in real time and brush to SSD.

Summary of the invention

(1) technical problems to be solved

The technical problem to be solved by the present invention is how to provide a kind of caching accelerated method of double storage controllers, realize After the failure of one storage control, the storage operation of data is taken over by another controller, i.e., caching system still can be after Continue and work in another controller, and guarantees that data storage service is not interrupted.

(2) technical solution

In order to solve the above-mentioned technical problems, the present invention provides a kind of dual control storage array solid state hard disks to cache acceleration side Method, comprising the following steps:

(1) it creation caching: initializes cache object respectively in the memory of two storage controls A, B, and establishes caching The caching of object indexes, and the corresponding caching index of each cache object, caching index is organized as a caching index tree, Cache object is stored in the caching index tree in non-leaf nodes and indexes key, corresponding cache object is found according to these key and exists Position on HDD；

The data cached mapping relations with HDD data of caching index record SSD, including HDD block data address, SSD Data block address, data block size, caching number and data access time；

Every four caching index organizations are a caching group, and each caching group is one piece of contiguous memory in memory, corresponding Unique caching group #；

The corresponding caching index tree of one rear end equipment, caching index tree are assigned with unique caching group #；

(2) caching read-write: traversal caching index updates the data access time of caching index, more for the caching of hit Another controller memory index tree is newly arrived, reads and writes data from the SSD data block address of index；For the caching in unnatural death, from HDD reads and writes data, and to memory index tree application available index, records the data to the corresponding SSD data block address of index；

(3) cache synchronization: the caching in storage control A memory is indexed by way of queue, is simultaneously sent to and deposits In the memory for storing up controller B, the caching of storage control B is indexed by way of queue, storage control A is simultaneously sent to In；

(4) it caches write-back: being executed by caching write-back thread, caching write-back thread statistics caching use space, caching uses Space is the sum of the size of data in all caching indexes, when caching use space higher than highest threshold value, caches write-back thread Start write-back buffer, by traversing the caching index tree, the caching group on index tree is cached described in poll, by each caching rope Draw the corresponding position that corresponding caching is written back to back-end logic volume, the caching index of write-back is by the synchronous setting of two storage controls Stop write-back buffer when spatial cache is lower than lowest threshold for available mode.

Preferably, in step (4), the corresponding caching of each caching index is written back to back-end logic volume by lru algorithm Corresponding position.

Preferably, the caching index tree tissue is a B+ tree.

(3) beneficial effect

The present invention provides a kind of caching accelerated method of double storage controllers, after realizing a storage control failure, The storage operation of data is taken over by another controller, i.e., caching system can still continue to work in another controller, and Guarantee that data storage service is not interrupted.Caching index tree real-time synchronization in two storage controls is in mirrored state.When one A storage control breaks down when failing, another controller can take in the storage on the fault point seamless pipe controller Business.The present invention both ensure that the efficient performance of SSD buffer memory, in turn ensure the high reliability of SSD buffer memory.This method can Applied to the business scenario in double storage controllers mixing storage array product, used simultaneously for SSD and HDD.It is total reducing On the basis of body possesses cost, while not only having ensure that reliability, but also the overall performance of product is improved.

Detailed description of the invention

Fig. 1 is caching index structure schematic diagram of the invention；

Fig. 2 is caching index tree of the invention and caching group structural schematic diagram；

Fig. 3 is that dual control of the invention caches accelerated method data flow figure.

Specific embodiment

To keep the purpose of the present invention, content and advantage clearer, with reference to the accompanying drawings and examples, to of the invention Specific embodiment is described in further detail.

As shown in Figure 1 to Figure 3, a kind of dual control storage array solid state hard disk caching accelerated method proposed by the present invention include with Lower step:

(1) it creation caching: initializes cache object respectively in the memory of two storage controls A, B, and establishes caching The caching of object indexes, and the corresponding caching index of each cache object, caching index is organized as a caching index tree, Cache object is stored in the caching index tree in non-leaf nodes and indexes key, corresponding cache object is found according to these key and exists Position on HDD creates following data structure in the present invention:

The caching index: the data cached mapping relations with HDD data of record SSD, including HDD block data address, SSD Data block address, data block size, caching number and data access time；

Caching group: every four caching index organizations are a caching group, and each caching group is in one piece continuous in memory It deposits, corresponding unique caching group #；

The caching index tree: the corresponding caching index tree of a rear end equipment, caching index tree tissue are a B+ Tree is assigned with unique caching group #；

(3) cache synchronization: the caching in storage control A memory is indexed by way of queue, is simultaneously sent to and deposits In the memory for storing up controller B, the caching of storage control B is indexed by way of queue, storage control A is simultaneously sent to In；In this way, the caching index tree in two storage control memories keeps realtime uniform；

(4) cache write-back: caching write-back is executed by special caching write-back thread, and write-back thread statistics caching uses empty Between, caching use space is that the sum of the size of data in all caching indexes is delayed when caching use space is higher than highest threshold value It is stored back to and writes thread and start write-back buffer, by traversing the caching index tree, caching group on caching index tree described in poll will Each caching index is corresponding to cache the corresponding position that back-end logic volume is written back to by lru algorithm, and the caching of write-back indexes quilt Two storage controls synchronize and are set as available mode, when spatial cache is lower than lowest threshold, stop write-back buffer.

The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art For member, without departing from the technical principles of the invention, several improvement and deformations can also be made, these improvement and deformations Also it should be regarded as protection scope of the present invention.

Claims

1. A dual-control storage array solid-state hard disk cache acceleration method, characterized in that, comprising the following steps:

(1) Create a cache: Initialize the cache objects in the memory of the two storage controllers A and B respectively, and establish a cache index of the cache object. Each cache object corresponds to a cache index, and the cache index is organized as a cache index tree. , store the cache object index key in the non-leaf node in the cache index tree, and find the position of the corresponding cache object on the HDD according to these keys;

The cache index records the mapping relationship between SSD cache data and HDD data, including HDD block data address, SSD data block address, data block size, cache number and data access time;

Every four cache indexes are organized into a cache group, and each cache group is a continuous memory in memory, corresponding to a unique cache group number;

A backend device corresponds to a cache index tree, and the cache index tree is assigned a unique cache group number;

(2) Cache read and write: traverse the cache index, update the data access time of the cache index for the hit cache, update the memory index tree of another controller, and read and write data from the indexed SSD data block address; for the non-hit cache , read and write data from the HDD, apply for an available index to the memory index tree, and record the data to the SSD data block address corresponding to the index;

(3) Cache synchronization: The cache index in the memory of storage controller A is synchronously sent to the memory of storage controller B through a queue, and the cache index of storage controller B is synchronously sent to the storage controller through a queue. in device A;

(4) Cache write-back: executed by the cache write-back thread. The cache write-back thread counts the cache usage space. The cache usage space is the sum of the data sizes in all cache indexes. When the cache usage space is higher than the highest threshold, the cache write-back is performed. The thread starts to write back the cache, by traversing the cache index tree, polling the cache group on the cache index tree, and writing back the cache corresponding to each cache index to the corresponding position of the back-end logical volume, the cache index of the write-back Synchronously set to available state by both storage controllers, stop write-back caching when the cache space falls below the minimum threshold.

2 . The method according to claim 1 , wherein, in step (4), the cache corresponding to each cache index is written back to the corresponding position of the back-end logical volume through the LRU algorithm. 3 .

3. The method of claim 1, wherein the cache index tree is organized as a B+ tree.

4. The method of claim 1, wherein the cache index trees in the two storage controllers are in a mirrored state.

5. The method of claim 1, wherein the method is applied to a dual storage controller hybrid storage array product.