CN117827097A

CN117827097A - Intelligent EC processing method and device

Info

Publication number: CN117827097A
Application number: CN202311708891.1A
Authority: CN
Inventors: 张宗全
Original assignee: Tianyi Cloud Technology Co Ltd
Current assignee: Tianyi Cloud Technology Co Ltd
Priority date: 2023-12-13
Filing date: 2023-12-13
Publication date: 2024-04-05

Abstract

The invention belongs to the technical field of computer storage, and particularly relates to an intelligent EC processing method and device. According to the invention, through the Chunk space allocation management, the writing operation on the database which is undergoing EC conversion can be effectively avoided, so that the problem of mutual exclusion of writing data is not concerned, the online EC mode conversion is realized, the protection mode of the database can be easily adjusted through the protection mode management based on the database and the control of a database read-write state machine, the intelligent EC is realized to dynamically adjust the EC proportion, when a single node expands to multiple nodes, the protection level can be changed online and improved from a disk level fault to a node level fault, even a rack level and data center level fault, and after a small-scale node expands to multiple nodes, the protection level can be changed online, the data storage efficiency is improved, and the storage cost per GB is reduced.

Description

Intelligent EC processing method and device

Technical Field

The invention belongs to the technical field of computer storage, and particularly relates to an intelligent EC processing method and device.

Background

EC (Erasure Code) is an encoding technique that can multiply n parts of original data by m parts and can restore the original data by any n parts of n+m parts. That is, if any m parts or less of data fail, the data can still be restored through the rest of data, which is a data redundancy technology currently used in the mainstream of distributed storage, and generally 4+2,8+3 or more 22+3 are used.

In the prior art, as the proportion of the data disk to the redundant disk increases, the storage efficiency is further improved, in distributed storage, the number of storage nodes is usually increased as required, and the traditional EC is not changed after EC or copy is created, so that the storage efficiency after capacity expansion cannot be improved, and the cost per TB of storage is affected.

Disclosure of Invention

The invention aims to provide an intelligent EC processing method and device, which can be used for improving storage reliability and usability from single-node EC to multi-node EC.

The technical scheme adopted by the invention is as follows:

an intelligent EC processing device comprises a Chunk allocation module, a PG view management module, a Pool of Pool storage resources, a Disk list and a Chunk Saver thread;

the Chunk allocation module is used for allocating a new data storage space for the additionally written storage system and writing new data of a user;

the PG view management module is used for acquiring a PG state, wherein the PG state comprises a read-only state, a read-write state and an uncertain state;

the Pool storage resource Pool comprises a plurality of PGs, and when the Pool storage resource Pool is created, the total number of PGs is synchronously determined;

the Disk list is used for listing all PGs, wherein each PG corresponds to a unique number;

the Chunk Saver threads are used for processing the Disk information, and each Chunk Saver thread is responsible for processing one or more Disk information.

In a preferred scheme, before the Chunk allocation module executes, the requirement of the upper layer application is acquired, and when the upper layer application needs to write data, a Chunk allocation request is generated and sent to the Chunk allocation module, and the Chunk allocation module allocates newly added Chunk space to the upper layer application.

In a preferred embodiment, the following specific process of allocating the newly added Chunk space to the upper layer application is performed by the Chunk allocation module:

acquiring a PG state corresponding to the newly added Chunk space and the PG space occupation amount;

and acquiring an allowable allocation threshold, comparing the allowable allocation threshold with the PG space occupation amount, and if the PG space occupation amount is larger than the allowable allocation threshold in the read-write state of the PG state, allocating Chunk from the PG to an upper layer application, otherwise, continuing to match the next PG or adding the PG.

In a preferred scheme, when the Pool storage resource Pool executes node capacity expansion, EC conversion is executed, a new PG is added, and a protection mode of the new PG is set as an EC mode to be converted, the Chunk distribution module applies for Chunk from the PG of the new EC mode to the upper layer application, then PG data arrangement is carried out on the original PG, and after the original PG is arranged into the new EC mode, the Chunk distribution module continues to apply for Chunk from the original PG to the upper layer application.

In a preferred scheme, the EC conversion comprises node capacity expansion and node capacity reduction, wherein the node capacity expansion gradually increases the EC proportion from node inside to node among nodes, and the node capacity reduction is to adjust the PG EC proportion.

In a preferred scheme, the steps for performing PG data arrangement on the original PG are as follows:

setting the original PG to be in a read-only state, adjusting the original PG to be in a new EC mode, and generating a PG reconstruction list;

acquiring all Chunk lists of data in the original PG, reading complete Chunk data, recoding the read complete Chunk data through a new PG EC protection mode, and writing the recoded data into the reconstructed Disk according to EC fragments until all Chunk data are subjected to EC recoding;

and updating the Disk list of the new PG to a reconstructed Disk list, changing the PG protection mode, simultaneously updating PG version information, and changing the tidied PG state to a read-write state.

In a preferred scheme, after the original PG is set to a read-only state, if a read request exists, EC decoding is performed from the original PG Disk list according to the original PG EC protection mode, and corresponding data in the Disk slices are read, otherwise, the original PG still performs EC mode conversion in the read-only state.

In a preferred scheme, after finishing PG data of the original PG, notifying a Disk in the original Disk list to recycle space occupied by the Chunk through a corresponding Chunk saver thread, and deleting Chunk information in the Disk;

the client does not change the PG state information, and a read request is sent to the Disk of the original PG Disk list for data reading, and the Disk of the original PG Disk list receives the request;

if the Chunk data is deleted, returning error information by the Chunk server thread corresponding to the Disk, and refreshing the PG view to the new Disk by the client to read Chunk fragment data;

if the Disk does not receive the information for deleting the Chunk, indicating that the Chunk exists, returning corresponding data.

In a preferred embodiment, the discs in the PG reconstruction list and the discs in the original PG disc list do not overlap each other.

The invention also provides an intelligent EC processing method which is applied to the intelligent EC processing device and comprises the following steps:

acquiring the requirement of an upper layer application, generating a Chunk allocation request when the upper layer application needs to write data, and allocating a newly added Chunk space from the PG in a read-write state according to the Chunk allocation request;

acquiring a capacity expansion node, executing EC conversion, updating a PG protection mode, changing the original PG state into a read-only state, and continuously receiving a read request;

performing data arrangement on the original PG, and after the original PG is arranged into a new EC mode, continuously applying for a Chunk from the original PG to an upper layer application by a Chunk distribution module;

and notifying the Disk in the original Disk list, recovering the space occupied by the Chunk, and deleting the Chunk information in the Disk.

The invention has the technical effects that:

according to the invention, through the Chunk space allocation management, the writing operation on the database which is undergoing EC conversion can be effectively avoided, so that the problem of mutual exclusion of writing data is not concerned, the online EC mode conversion is realized, the protection mode of the database can be easily adjusted through the protection mode management based on the database and the control of a database read-write state machine, the intelligent EC is realized to dynamically adjust the EC proportion, when a single node expands to multiple nodes, the protection level can be changed online and improved from a disk level fault to a node level fault, even a rack level and data center level fault, and after a small-scale node expands to multiple nodes, the protection level can be changed online, the data storage efficiency is improved, and the storage cost per GB is reduced.

Drawings

FIG. 1 is a block diagram of an apparatus provided by the present invention;

fig. 2 is an exemplary diagram of an original PG data arrangement flow provided by the present invention;

fig. 3 is a flow chart of a method provided by the present invention.

Detailed Description

In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.

Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one preferred embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.

Referring to fig. 1, the present invention provides an intelligent EC processing apparatus, which includes a Chunk allocation module, a PG view management module, a Pool storage resource Pool, a Disk list, and a Chunk Saver thread;

In the invention, the EC technology is the data redundancy technology mainly used at present in distributed storage, and is generally used as 22+3 with the proportion of 4+2,8+3 or more, in the EC technology, as the proportion of data discs to redundant discs increases, the storage efficiency is further improved, in distributed storage, the number of storage nodes is generally increased as required, but the traditional EC is not changed after EC or copy is created, so that the storage efficiency after expansion cannot be improved, the cost of each TB is influenced, in the embodiment, a Chunk allocation module, a PG view management module, a Pool storage resource Pool, a Disk list and a Chunk save thread are arranged, the Chunk allocation module has high intelligence and automation, can intelligently allocate new data storage space according to the load and the requirement of a storage system, thereby ensuring that written user data can be stored and managed properly, the module effectively improves the utilization rate of the storage space through precise calculation and algorithm, can flexibly adjust the allocation strategy of the storage space to adapt to different user requirements and data types, the PG view management module is a powerful and efficient management tool which can acquire and monitor the states of PG in real time, the states comprise a read-only state, a read-write state and an uncertain state, each state represents different functions and purposes of PG, for example, the read-only state indicates that PG can only read data, the read-write state indicates that PG can read and write data, the uncertain state indicates that PG state may exist or be unstable, further checking and confirmation are needed, the Pool is a comprehensive storage solution, the PG is determined when the Pool is created, and each PG has a unique number, the design ensures that the Pool storage resource Pool has high expandability and flexibility, PGs can be easily added or deleted to meet the changing storage requirement, a Disk list is an important component part which lists all PGs in the Pool storage resource Pool, the list not only provides quick access to each PG, but also enables a user to easily know and manage each element in the Pool storage resource Pool, a Chunk Saver thread is a key component responsible for processing Disk information, and each Chunk Saver thread is responsible for processing one or more Disk information, so that the integrity and consistency of data are ensured.

In a preferred embodiment, before the Chunk allocation module executes, the requirements of the upper layer application are obtained, and when the upper layer application needs to write data, a Chunk allocation request is generated and sent to the Chunk allocation module, and the Chunk allocation module allocates newly added Chunk space to the upper layer application.

In this embodiment, before executing the Chunk allocation module, the requirement of the upper layer application needs to be acquired first, specifically, when the upper layer application needs to write data, a Chunk allocation request is generated and sent to the Chunk allocation module, and the module allocates newly added Chunk space for the upper layer application immediately.

Secondly, the concrete process of distributing the newly added Chunk space to the upper layer application by the Chunk distribution module is as follows:

In this embodiment, before allocating newly added Chunk space to an upper layer application, it is necessary to acquire PG states of each PG in advance, where only PGs in a read-write state and PG space occupation amount greater than an allowable allocation threshold are considered to be allocatable, otherwise Chunk space may be allocated to the upper layer application by adding PGs.

And secondly, when the Pool storage resource Pool executes node capacity expansion, executing EC conversion, adding a new PG, setting a protection mode of the new PG as an EC mode needing conversion, applying a Chunk from the PG of the new EC mode to an upper layer application by the Chunk distribution module, performing PG data arrangement on the original PG, and continuing applying the Chunk from the original PG to the upper layer application by the Chunk distribution module after the original PG is arranged into the new EC mode.

In this embodiment, when the EC conversion is executed, when a single is supported, the inter-disk EC is used in the node to implement disk level fault processing and recovery, when the number of nodes is 2 or 3, the inter-node EC is used to implement inter-node data reliability management by using inter-node copies or folded ECs to implement node and fault processing and recovery, when the number of nodes is greater than 3, an EC mode with higher storage efficiency is adopted to implement inter-node data reliability management, the storage space utilization rate is improved, the EC can gradually increase the number of EC disks from 4+2 to 20+3 according to the number of nodes, the data redundancy efficiency is improved, by configuring a default number of redundant data copies, when the number of nodes is greater than the number of redundant copies +1, the system automatically sorts the data which does not meet the reliability requirement, for example, from the single-node EC to the inter-node EC (3 copies or 4+2:1ec can also be sorted into EC with higher storage efficiency, for example, 16+3) can also be set, when the number of nodes is set to have the availability priority, the number of available is less than the set, the number of redundant data can be gradually increased from 4+2 to 20+3, the number of redundant data can still be distributed to the original PG, and all PG can be distributed from the PG system to the full-scale, and the PG can be distributed according to the whole PG, and the whole PG is distributed according to the whole data, and the whole process can be distributed.

In a preferred embodiment, the EC transformation includes node expansion and node contraction, the node expansion gradually increases the EC ratio from node to node, and the node contraction is to adjust the PG EC ratio.

In this embodiment, when the node capacity expansion is performed, taking the total PG number as 2048 as an example, when a single node is used, PGs 1 to 16 are used, when a PG redundancy mode is set as EC:4+2,3 nodes in the node, PGs 17 to 64 are newly used, the PG redundancy mode is set as 3 pairs (4+2: 1) of this mode, and the new application data uses PGs 17 to 64, thereby ensuring that the new data can all meet the reliability requirement. Simultaneously setting PG 1-16 to be in a read-only state, when upgrading to N nodes, newly using PG 65-1024, setting PG redundancy mode to be an inter-node EC mode, and using PG 65-1024 for new application data so as to ensure that the new data can meet the reliability requirement, and improving the data storage efficiency;

when the node capacity reduction is executed, taking 10+2 as an example, firstly, the original 10+2PG state is adjusted to be in a read-only state, PGs 1025-1224 are additionally used, a PG redundancy mode is set to be (8+2), PG 1025-1224 are used for new application data, the PG data of the original 10+2 are sorted, the conversion is carried out to 8+2EC, and the PG state is set to be readable and writable after finishing the sorting.

Next, as shown in fig. 2, the steps for performing PG data arrangement on the original PG are as follows:

In this embodiment, when the original PG is subjected to PG data arrangement, the PG EC protection mode is changed, the PG is set to a read-only state, and the EC protection mode that is adjusted according to the need is used to generate a PG reconstruction list, the discs in the PG reconstruction list and the discs in the original PG disc list are not repeated, as in PG126 in fig. 2, when a read request is made, EC decoding is performed according to mode 8+2 from the disc list of PG126, corresponding data in the Disk slices is read, all the Chunk lists (which can be obtained one by one) with data in the PG are obtained, then the complete Chunk data is read, the new encoded data is encoded through the new protection mode 4+2, the EC slices are written into the reconstructed Disk, and finally the arranged PG state is changed to the read-write state.

And then, after the original PG is set to be in a read-only state, if a read request exists, EC decoding is carried out from the original PG Disk list according to an original PG EC protection mode, and corresponding data in the Disk fragments are read, otherwise, the original PG still executes EC mode conversion in the read-only state.

In this embodiment, the original PG is set to a read-only state, and still can receive a read request, at this time, EC decoding is performed from the original PG Disk list according to the original PG EC protection mode, and then the corresponding data in the Disk slices is read, and the PG in this state does not receive a write request.

In a preferred embodiment, after finishing the PG data of the original PG, notifying a Disk in the original Disk list to recycle the space occupied by the Chunk by a corresponding Chunk saver thread, and deleting Chunk information in the Disk;

In this embodiment, after the PG data of the original PG is finished, there may be a case that the client does not change the PG state information in time, the read request may be sent to the Disk of the original PG Disk list to read the data, after the Disk of the original PG Disk list receives the request, there may be two cases, that is, one of the two cases, the Chunk data is deleted, at this time, the Disk (corresponding Chunk server thread) may return an error message, the client refreshes the PG view to a new Disk to read Chunk data, and the second case, the Disk does not receive the information of deleting the Chunk, and if the Chunk exists, the corresponding data is returned, and since the data is not rewritten, the data correctness is guaranteed.

Referring to fig. 3, the present invention further provides an intelligent EC processing method, which is applied to the intelligent EC processing device, and includes:

s1, acquiring the requirement of an upper layer application, generating a Chunk allocation request when the upper layer application needs to write data, and allocating a newly added Chunk space from PG in a read-write state according to the Chunk allocation request;

s2, acquiring a capacity expansion node, executing EC conversion, updating a PG protection mode, changing the original PG state into a read-only state, and continuously receiving a read request;

s3, performing data arrangement on the original PG, and after the original PG is arranged into a new EC mode, continuously applying for a Chunk from the original PG to an upper layer application by a Chunk distribution module;

s4, notifying the Disk in the original Disk list, recovering the space occupied by the Chunk, and deleting the Chunk information in the Disk.

As described in the above steps S1-S4, the requirements of the upper layer application are firstly obtained to ensure that the write-in request of the data is satisfied, when the upper layer application needs to write in the data, a Chunk allocation request is generated to ensure that enough space is stored in the data, according to the Chunk allocation request, newly increased Chunk space is allocated from the PG in the read-write state to ensure the storage and management of the data, then the capacity-expansion node is obtained, and EC conversion is executed to ensure the stable storage of the data, meanwhile, the PG protection mode is updated to protect the safety and integrity of the data, then the original PG state is changed into the read-only state to prevent the data from being modified accidentally, and meanwhile, the read request is continuously received to satisfy the read-in requirement of the data by the user, then the original PG is subjected to data arrangement according to a new EC mode to ensure that the data is better managed and stored, after the original PG is arranged into the new EC mode, the Chunk allocation module continues to apply for the Chunk to the upper layer application from the original PG to ensure the storage and management stability and reliability of the data, and finally the disc in the original disc occupation list is notified, the occupied space of the Chunk space is prevented from being wasted, and the data is prevented from being deleted and the disc is deleted and the storage information is ensured.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, apparatus, article or method that comprises the element.

The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention. Structures, devices and methods of operation not specifically described and illustrated herein, unless otherwise indicated and limited, are implemented according to conventional means in the art.

Claims

1. The utility model provides an intelligent EC processing device, includes Chunk distribution module, PG view management module, pool storage resource Pool, disk list and Chunk Saver thread, its characterized in that:

2. A smart EC processing device according to claim 1, wherein: and before the Chunk allocation module executes, acquiring the requirement of an upper layer application, generating a Chunk allocation request when the upper layer application needs to write data, and sending the Chunk allocation request to the Chunk allocation module, wherein the Chunk allocation module allocates newly added Chunk space to the upper layer application.

3. A smart EC processing device according to claim 2, wherein: the specific process of the Chunk allocation module for allocating newly added Chunk space to the upper layer application is as follows:

4. A smart EC processing device according to claim 1, wherein: and when the Pool storage resource Pool executes node capacity expansion, executing EC conversion, adding a new PG, setting a protection mode of the new PG as an EC mode needing conversion, applying for a Chunk from the PG of the new EC mode to an upper layer application by the Chunk distribution module, performing PG data arrangement on the original PG, and continuing applying for the Chunk from the original PG to the upper layer application by the Chunk distribution module after the original PG is arranged into the new EC mode.

5. The smart EC processing device of claim 4, wherein: the EC conversion comprises node capacity expansion and node capacity reduction, wherein the node capacity expansion gradually increases the EC proportion from inside to between nodes, and the node capacity reduction is to adjust the PG EC proportion.

6. The smart EC processing device of claim 4, wherein: the steps for the PG data arrangement of the original PG are as follows:

7. The smart EC processing device of claim 5, wherein: after the original PG is set to be in a read-only state, if a read request exists, EC decoding is carried out from the original PG Disk list according to an original PG EC protection mode, and corresponding data in the Disk fragments are read, otherwise, the original PG still executes EC mode conversion in the read-only state.

8. The smart EC processing device of claim 5, wherein: after the PG data of the original PG is finished, notifying a Disk in the original Disk list to recover the space occupied by the Chunk through a corresponding Chunk saver thread, and deleting Chunk information in the Disk;

9. The smart EC processing device of claim 5, wherein: the discs in the PG reconstruction list and the discs in the original PG disc list do not overlap each other.

10. A smart EC processing method applied to the smart EC processing apparatus of any of claims 1 to 9, characterized in that: comprising the following steps: