WO2015067004A9

WO2015067004A9 - Concurrent access request processing method and device

Info

Publication number: WO2015067004A9
Application number: PCT/CN2014/075558
Authority: WO
Inventors: 童燕群; 李成林
Original assignee: 华为技术有限公司
Priority date: 2013-11-07
Filing date: 2014-04-17
Publication date: 2015-09-03
Also published as: WO2015067004A1; CN103634374B; CN103634374A

Abstract

Disclosed are a concurrent access request processing method and device to avoid concurrent access conflict. The method comprises: receiving at least two concurrent access requests for the same data resource and sorting the at least two concurrent access requests; and accessing the same data resource sequentially according to the sorted concurrent access requests. The present invention ensures that there is only the data resource corresponding to one request access, thereby avoiding concurrent access conflict.

Description

Method and apparatus for processing concurrent access requests. The present application claims priority to Chinese patent application filed on November 7, 2013, the Chinese Patent Office, Application No. 201310549721.3, the name of the invention, and the processing method and apparatus of a concurrent access request. The entire contents thereof are incorporated herein by reference.

The present invention relates to the field of computer data storage technologies, and in particular, to a method and apparatus for processing concurrent access requests. Background technique

The object storage technology based on HTTP (Hyper Text Transfer Protocol) is rapidly developing, and the object storage technology based on the two-layer business model of container Container and object Object is more and more widely used.

A container can be understood as a special top-level directory. An object can be a file or a directory, and the object belongs to the container. Usually, user data is stored in a container in an object mode, using a technical architecture in which an upper layer application establishes object storage on the underlying distributed storage engine. The number of objects in the container is unrestricted. When there are many objects in the container, there will be a very large index table between the container and the object. Ordinary stand-alone databases can no longer meet the storage requirements. Therefore, the storage engine based on weak consistency is generally maintained in a B-tree structure.

Figure 1 shows a simple B-tree structure; Figure 2 shows an object storage system based on the underlying distributed storage engine architecture. The sub-blocks N1, N2, N3, ... of the B-tree structure in Fig. 1 can be stored in the figure.

On one or more of the child nodes 1, 2, 3, ... in the distributed storage engine, for example, the child data block N1 may be stored on the child node 2, the child node 4, and the child node 6. When writing a record to a B-tree structure, the process of "read-modify-write" is generally employed. In Figure 2, based on the object storage system set up by the underlying distributed storage engine, it is reflected that the client initiates an access request to the data resource when accessing After the data resource is obtained, the data resource is modified and then written back to the access storage engine. The initiating access request process is: the client based on the HTTP protocol sends an access request for accessing the data resource, and the upper layer application analyzes the metadata of the related container and the object and the data resource to be accessed from the received access request, and then The corresponding engine access proxy of the upper application requests data resources from the underlying distributed storage engine.

When object storage is performed, when there are multiple clients adding objects to the same container at the same time, the sub-blocks in the B-tree structure become hot spots. For example, there may be multiple clients that need to write the sub-block N1 into the B-tree at the same time, which will cause the sub-block N1 in the B-tree structure to become a hot spot. In the object storage system based on the underlying distributed storage engine erected in FIG. 2, a concurrent access request is initiated for multiple clients, so that multiple engine access agents simultaneously request access to the child nodes of the storage engine that store the sub-block N1. , causing concurrent access violations of the underlying storage engine. As shown in FIG. 3, if the sub-block N1 is stored on the sub-node 6, the upper-layer applications APP1, APP2, and APP3 will simultaneously request access to the sub-node 6 through their respective engine access agents.

When the storage engine based on the weak consistency uses the B-tree structure to maintain the index list, when the underlying storage engine has a concurrent access conflict, the underlying distributed storage engine returns a data conflict response to the upper layer application, and the upper layer application according to the specific service. Choose to rewrite. However, when there are multiple concurrency conflicts, it will affect the write performance of the B-tree structure, and even cause the data to be rewritten, eventually resulting in the loss of objects in the container.

Therefore, in an object storage system based on the underlying distributed storage engine, it is important to avoid concurrent access conflicts when concurrent access requests are made to data resources that become hotspots. Summary of the invention

Embodiments of the present invention provide a method and apparatus for processing concurrent access requests to avoid concurrent access conflicts.

In a first aspect, a method for processing a concurrent access request is provided, including:

Receiving at least two concurrent access requests for the same data resource, and sorting the at least two concurrent access requests; The same data resource is accessed in turn according to the sorted concurrent access request.

In conjunction with the first aspect, in a first possible implementation, the concurrent access request is sorted, including:

The at least two concurrent access requests are ordered in an engine access proxy.

In conjunction with the first aspect, in a second possible implementation, the sorting the at least two concurrent access requests includes:

Corresponding to a hash key value for each data resource on the storage engine to form a hash space;

Divide the hash space into N sub-hash spaces, where N is the number of engine access agents in the storage system;

Allocating N sub-hash spaces to N engine access agents, so that each engine access agent is assigned a sub-hash space, and different sub-hash spaces allocated by different engine access agents are different;

Determining an engine accessing agent that sorts the at least two concurrent access requests according to a sub-hash space to which the hash value of the same data resource accessed by the at least two concurrent access requests belongs, to obtain a sorting engine accessing the proxy;

An access request that is not on the sorting engine access agent is routed to the sorting engine access agent, and the sorting engine accesses the agent to sort the at least two concurrent access requests.

With the second possible implementation of the first aspect, in a third possible implementation, after the hash space is divided into N sub-hash spaces, the method further includes:

Monitor the number of engine access agents and repartition the hash space as the number of engine access agents changes.

In conjunction with the first possible implementation of the first aspect, in a fourth possible implementation, the same data resource is cached in an engine access proxy that sorts the concurrent access requests.

In a second aspect, a processing apparatus for concurrent access request is provided, including a receiving unit, a sorting unit, and an access unit, where

The receiving unit is configured to receive at least two concurrent access requests for the same data resource, and transmit the at least two concurrent access requests to the sorting unit; The sorting unit is configured to receive the at least two concurrent access requests transmitted by the receiving unit, and sort the at least two concurrent access requests, and transmit the sorted concurrent access request to the access unit;

The access unit is configured to receive the sorted concurrent access request transmitted by the sorting unit, and sequentially access the same data resource according to the sorted concurrent access request.

With reference to the second aspect, in a first possible implementation, the sorting unit is specifically configured to: in an engine access proxy, sort the at least two concurrent access requests.

With reference to the second aspect, in a second possible implementation manner, the sorting unit is specifically configured to: associate each data resource on the storage engine with a hash key value to form a hash space;

Divide the hash space into a sub-hash space, where Ν is the number of engine access agents in the storage system;

Assigning a sub-hash space to each of the engine access agents, so that each engine access agent is assigned a sub-hash space, and different sub-hash spaces are assigned by different engine access agents;

In conjunction with the second possible implementation of the second aspect, in a third possible implementation, the sorting unit is further configured to:

In conjunction with the first possible implementation of the second aspect, in a fourth possible implementation, a cache unit is further included, where

The cache unit is configured to cache the same data resource in an engine access proxy that sorts the concurrent access requests. The concurrent access request processing method provided by the first aspect of the present invention and the concurrent access request processing apparatus provided by the second aspect, after receiving at least two concurrent access requests for the same data resource, sorting the at least two concurrent access requests According to the sorted concurrent access request, sequentially accessing the same data resource can ensure that only one request accesses the corresponding data resource at the same time, thereby avoiding concurrent access conflicts. DRAWINGS

1 is a schematic diagram of an organization structure of an index table based on a B-tree structure in the prior art;

2 is a schematic diagram of an architecture of an object storage system based on an underlying distributed storage engine in the prior art;

FIG. 3 is a schematic diagram of an access violation occurring in a concurrent access request in the prior art; FIG.

4 is a schematic flowchart of a method for processing a concurrent access request according to an embodiment of the present invention;

FIG. 5 is a flowchart of a method for scheduling concurrent access requests according to an embodiment of the present invention;

6 is a schematic diagram of hash space division in an embodiment of the present invention;

7 is a schematic diagram of a process of sorting concurrent access requests in an embodiment of the present invention;

8 is a schematic structural diagram of a concurrent access request processing apparatus according to an embodiment of the present invention; FIG. 9 is another schematic structural diagram of a concurrent access request processing apparatus according to an embodiment of the present invention;

FIG. 11 is still another schematic diagram of a controller configuration according to an embodiment of the present invention;

FIG. 12 is still another schematic diagram of a controller configuration according to an embodiment of the present invention. detailed description

The technical solutions in the embodiments of the present invention will be clearly and completely described in the following with reference to the accompanying drawings in the embodiments. It is apparent that the described embodiments are only a part of the embodiments of the invention, not all of them.

Based on the object storage system set up by the underlying distributed storage engine, different clients may initiate concurrent access requests to the same data resource, and the upper layer application may receive at least two concurrent data resources. Access request, the at least two concurrent access requests are processed by different ΑΡΡ (upper application) and engine access agent. In the embodiment of the present invention, the at least two concurrent access requests are processed in the upper layer application to ensure that only one request accesses the corresponding data resource at the same time, thereby preventing concurrent access conflicts.

Embodiment 1

A first embodiment of the present invention provides a method for processing a concurrent access request, as shown in FIG. 4, including:

5101: Receive at least two concurrent access requests for the same data resource.

Specifically, for a data resource with hotspot access, in an object storage system based on the underlying distributed storage engine, an access request for the data resource is initiated by different clients based on the HTTP protocol, that is, at least the same data resource exists. Two concurrent access requests. The upper application receives at least two concurrent access requests for the same data resource.

5102: Sort the received at least two concurrent access requests.

Specifically, in the embodiment of the present invention, at least two concurrent access requests of the same data resource received by the upper layer application are sorted, and may be directly sorted on different engine access agents that request data resources of the underlying storage engine, or may be different. The access request on the engine access proxy is routed to an engine access proxy for sorting. The embodiment of the present invention preferably routes access requests on different engine access agents to an engine access agent for sorting, so that when the concurrent access requests are sorted, no state query is required between different engine access agents.

5103: Access the data resources in turn according to the sorted concurrent access request.

In the embodiment of the present invention, at least two concurrent access requests for accessing the same data resource are sorted. According to the sorted concurrent access request, accessing the data resource in turn can ensure that only one request accesses the data resource at a time, thereby avoiding concurrent access conflicts.

Embodiment 2

The second embodiment of the present invention will be described in detail in conjunction with the actual application for the concurrent access request processing method involved in the first embodiment.

For object storage technologies based on two-tier business models of containers and objects, when there is more than one When a client adds an object to a container at the same time, it causes the sub-block data of the index relation table to become a hot spot, that is, there is more than one access request. The access table between the container and the object is requested to be accessed. In the embodiment of the present invention, the concurrent access requests of the same index table are sorted to ensure that the same index table has only one read and write operation at the same time, thereby avoiding concurrent access conflicts.

The process of sorting the access request in the embodiment of the present invention is described in detail. For the processing steps of the concurrent access request, refer to the first embodiment, and details are not described herein again.

In an embodiment of the present invention, in an engine access proxy, at least two concurrent access requests of the same data resource are sorted, and access requests on different engine access agents are routed to an engine access proxy for sorting, so that concurrent access is performed. When requesting a sort, there is no need for a different state to access the status query between the agents.

In the embodiment of the present invention, in an engine accessing agent, the process of sorting at least two concurrent access requests is as shown in FIG. 5, including:

5201: Corresponding to a hash key value for each data resource on the storage engine to form a hash space. In the embodiment of the present invention, when there is more than one access request requesting access to the index table between the container and the object, the data resource on the storage engine is hashed according to a unified hash algorithm between the upper layer applications. Each data resource corresponds to a hash key value and constitutes a hash space. The hash space is a one-dimensional space that is sufficiently large for the number of APPs applied to the upper layer.

5202: The hash space in the partition S201 is N sub-hash spaces.

In the embodiment of the present invention, the number of divided sub-hash spaces is the same as the number of engine access agents corresponding to the upper layer application. If the number of the upper layer application is N, the hash space is divided into N parts, and N sub-hash spaces are obtained. For example, in the embodiment of the present invention, the upper layer application includes APP1, APP2, and APP3, and the hash space partitioning process is shown in FIG. 6. Shown.

Further, when performing hash space partitioning in the embodiment of the present invention, the partitioning may be performed by using an equal division method, or may be performed by using an unequal division method, as long as the number of divided sub-hash spaces is corresponding to the engine accessing agent corresponding to the upper layer application. The number is the same. For example, in the embodiment of the present invention, the number of upper layer applications is N, and if the hash space can be equally divided into N parts, in the embodiment of the present invention, the hash may be equally divided. The space is divided into N parts, and a sub-hash space with the same hash space is obtained. If the hash space cannot be equally divided, the hash space can be divided into N by using the unequal division method in the embodiment of the present invention. - 1 sub-hash space with equal hash space, divide the remaining hash space into 1 sub-hash space, and finally get a sub-hash space.

5203: Assign a sub-hash space to the engine access agent.

In the embodiment of the present invention, a simple cluster can be established between the engine access agents of the upper layer application, the agent number is accessed for each engine, and a sub-hash space is allocated for each engine access agent's own sequence number. The sub-hash space divided in S202 is allocated to one engine access agent, and each engine access agent allocates a sub-hash space, and different sub-hash spaces allocated by different engine access agents are different.

5204: Determine a hash value of at least two data resources requested to access, and determine an engine access proxy that sorts at least two concurrent access requests according to a sub-hash space allocated to the sub-engine access proxy, and obtain a sorting engine access proxy. .

After the data resource is hashed in the embodiment of the present invention, each data resource corresponds to a hash value, and the hash value belongs to a sub-hash space obtained by dividing in S202. And each hash space is assigned to an engine access agent. Therefore, the engine access agent that sorts the at least two concurrent access requests may be determined according to the sub-hash space to which the hash value of the same data resource accessed by the at least two concurrent access requests belongs. The determined engine access agent is referred to as a sorting engine access agent through which the at least two concurrent access requests are sorted.

5205: Forward concurrent access requests that are not on the sort engine access agent to the sorting engine access agent.

Each access request in the embodiment of the present invention is sent to the upper application ,, so different access requests will be on different engine access agents. In the embodiment of the present invention, after determining that the sorting engine accesses the proxy in S204, the access request that is not on the determined sorting engine accessing proxy is routed to the sorting engine accessing proxy, and the sorting engine accesses the proxy, and routes to the proxy All concurrent access requests are sorted. For example, there are currently 3 concurrent access requests that need to be accessed at the storage engine. The hash value obtained by hashing the data resource on the node 6 and storing the data resource on the engine sub-node 6 belongs to the engine access proxy corresponding to APP1. Then, the access request on the engine access proxy corresponding to APP2 and APP3 is respectively routed to the engine access proxy corresponding to APP1. The three concurrent access requests are sorted by the engine access agent corresponding to APP1, as shown in FIG.

In the embodiment of the present invention, through the foregoing processing manner of the concurrent access request, the request for the index table of the same container is routed to the engine access proxy of the same application for sorting processing, and sequentially accessed according to the sorted concurrent access request. Index tables ensure that there are no access violations, thus avoiding data conflicts with the underlying distributed storage engine.

Further, in the embodiment of the present invention, after the sub-hash space is divided by S202, the number of accessing agents of the engine is monitored. When the number of engine access agents changes, the hash space is re-divided to accommodate the case where the upper application node exits or newly joins.

Further, since the data writing in the index table sub-block of the B-tree structure is performed on the basis of the last completed data. Therefore, in the embodiment of the present invention, when an engine access proxy is used to sort concurrent access requests, the read/write cache may be added to the engine access proxy that sorts the concurrent access requests, and the data resources are cached to improve the data resources. Access speed.

Embodiment 3

Based on the concurrent access request processing method provided by the first embodiment and the second embodiment, the embodiment of the present invention provides a concurrent access request processing device. As shown in FIG. 8, the concurrent access request processing device provided by the embodiment of the present invention includes a receiving unit. 801, a sorting unit 802 and an access unit 803, wherein

The receiving unit 801 is configured to receive at least two concurrent access requests for the same data resource, and transmit at least two concurrent access requests to the sorting unit 802.

The sorting unit 802 is configured to receive at least two concurrent access requests transmitted by the receiving unit 801, and sort the at least two concurrent access requests, and transmit the sorted concurrent access requests to the access unit 803;

The access unit 803 is configured to receive the sorted concurrent access request transmitted by the sorting unit 802, and sequentially access the same data resource according to the sorted concurrent access request. Specifically, in the embodiment of the present invention, the sorting unit 802 is configured to sort at least two concurrent access requests in an engine access proxy.

The sorting unit 802 is specifically configured to:

Determining an engine access proxy that sorts at least two concurrent access requests according to a sub-hash space to which the hash value of the same data resource accessed by the at least two concurrent access requests belongs, to obtain a sorting engine accessing the proxy;

An access request that is not on the sort engine access proxy is routed to the sort engine access proxy, and the sort engine accesses the proxy to sort at least two concurrent access requests.

Further, the sorting unit 802 in the embodiment of the present invention is further configured to:

The concurrent access request processing apparatus provided by the embodiment of the present invention further includes a cache unit 804, as shown in FIG. 9, wherein the cache unit 804 is configured to cache the same data resource in an engine access proxy that sorts concurrent access requests.

The concurrent access request processing apparatus provided by the embodiment of the present invention, when there are at least two concurrent access requests for the same data resource, sorts the at least two concurrent access requests, and sequentially accesses the same data according to the sorted concurrent access request. The resource can ensure that only one request accesses the corresponding data resource at a time, thereby avoiding concurrent access conflicts.

The above-mentioned concurrent access request processing device provided by the embodiment of the present invention may be an independent component, or may be integrated into other components. For example, the concurrent access request processing device provided by the embodiment of the present invention may be an engine access proxy, or may be New components integrated into the engine access agent. It should be noted that the function implementation and the interaction manner of each module/unit of the concurrent access request processing apparatus in the embodiment of the present invention may be further referred to the description of the related method embodiments.

Embodiment 4

The embodiment of the present invention provides a controller, which can be applied to an object storage service based on a container and an object two-layer service model, as shown in FIG. 10, in accordance with an embodiment of the present invention. The controller includes: a processor 1001 and an I/O interface 1002, wherein the processor 1001 is configured to receive at least two concurrent access requests for the same data resource, and receive the received at least two concurrent access requests. Sorting, and transmitting the sorted concurrent access request to the I/O interface 1002;

The I/O interface 1002 is configured to receive the sorted concurrent access request transmitted by the processor 1001, and output the sorted concurrent access request.

Further, the processor 1001 is configured to sort at least two concurrent access requests in an engine access agent.

The processor 1001 is specifically configured to: associate each data resource on the storage engine with a hash key value to form a hash space; divide the hash space into a sub-hash space, where Ν is the number of engine access agents; A sub-hash space is allocated to each of the engine access agents, so that each engine access agent is assigned a sub-hash space, and different sub-hash spaces are assigned by different engine access agents; accesses according to at least two concurrent access requests The sub-hash space to which the hash value of the same data resource belongs, determine the engine access proxy that sorts at least two concurrent access requests, obtain the sorting engine access proxy, and route the access request that is not in the sorting engine access proxy to the sorting On the engine access agent, the sorting engine accesses the agent and sorts at least two concurrent access requests.

Further, the controller in the embodiment of the present invention further includes a monitor 1003. As shown in FIG. 11, the monitor 1003 monitors the number of engine access agents, and sends a re-division to the processor 1001 when the number of engine access agents changes. Hash space instructions.

Further, in the embodiment of the present invention, the controller further includes a buffer 1004. As shown in FIG. 12, the buffer 1004 is used in an engine access proxy for sorting concurrent access requests by the processor 1001. Cache the same data resource.

The controller provided by the embodiment of the present invention, when there are at least two concurrent access requests for the same data resource, sort the at least two concurrent access requests, and sequentially access the same data resource according to the sorted concurrent access request. It is guaranteed that only one request for accessing the corresponding data resource at the same time can avoid concurrent access conflicts.

It is apparent that those skilled in the art can make various modifications and variations to the invention without departing from the spirit and scope of the invention. Thus, it is intended that the present invention cover the modifications and the modifications of the

Claims

claims

1. A method for processing concurrent access requests, which is characterized by including:

Receive at least two concurrent access requests for the same data resource, and sort the at least two concurrent access requests;

The same data resource is accessed sequentially according to the sorted concurrent access requests.

2. The method of claim 1, characterized in that sorting the concurrent access requests includes:

In an engine access agent, the at least two concurrent access requests are sequenced.

3. The method of claim 1, wherein said sorting the at least two concurrent access requests includes:

Each data resource on the storage engine corresponds to a hash key value to form a hash space;

Allocate N sub-hash spaces to N engine access agents, so that each engine access agent is assigned a sub-hash space, and different engine access agents are assigned different sub-hash spaces;

According to the sub-hash space to which the hash values of the same data resource accessed by the at least two concurrent access requests belong, determine the engine access agent that sorts the at least two concurrent access requests, and obtain the sorting engine access agent;

Access requests that are not on the sorting engine access proxy are routed to the sorting engine access proxy, and the sorting engine access proxy sorts the at least two concurrent access requests.

4. The method of claim 3, wherein after dividing the hash space into N sub-hash spaces, the method further includes:

Monitor the number of engine access proxies. When the number of engine access proxies changes, the hash space is re-divided.

5. The method of claim 2, wherein the method further includes:

In the engine access proxy that sorts the concurrent access requests, the same data resource is cached. source.

6. A device for processing concurrent access requests, characterized by including a receiving unit, a sorting unit and an access unit, wherein,

The receiving unit is configured to receive at least two concurrent access requests for the same data resource and transmit the at least two concurrent access requests to the sorting unit;

The sorting unit is configured to receive the at least two concurrent access requests transmitted by the receiving unit, sort the at least two concurrent access requests, and transmit the sorted concurrent access requests to the access unit;

The access unit is configured to receive the sorted concurrent access requests transmitted by the sorting unit, and to access the same data resource in sequence according to the sorted concurrent access requests.

7. The processing device according to claim 6, wherein the sorting unit is specifically used for:

8. The processing device according to claim 6, wherein the sorting unit is specifically used for:

9. The processing device according to claim 8, wherein the sorting unit is also used for: Monitor the number of engine access proxies, and re-divide the hash space when the number of engine access proxies changes.

10. The processing device according to claim 7, further comprising a cache unit, wherein the cache unit is used to cache the same data in the engine access agent that sorts the concurrent access requests. resource.