CN112256438B

CN112256438B - Load balancing control method and device, storage medium and electronic equipment

Info

Publication number: CN112256438B
Application number: CN202011273203.XA
Authority: CN
Inventors: 薛彦涛; 张智
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-06-28
Filing date: 2020-06-28
Publication date: 2021-06-25
Anticipated expiration: 2040-06-28
Also published as: CN112256438A; CN111666159A; CN111666159B

Abstract

The invention discloses a load balancing control method and device, a storage medium and electronic equipment. The method comprises the following steps: acquiring an average load value of load values of each metadata server in the metadata server cluster and a current load value of a current metadata server in the metadata server cluster; determining that the current metadata server is in a relative overload state under the condition that the current load value of the current metadata server is larger than the average load value; comparing the current load value with the load upper limit value configured for the current metadata server under the condition that the current metadata server is in a relative overload state; determining that the current metadata server is in an overload state under the condition that the current load value is greater than the load upper limit value configured for the current metadata server; and carrying out metadata migration on the current metadata server in the overload state. The invention solves the problem of object storage resource waste in the load balancing control method provided by the related technology.

Description

Load balancing control method and device, storage medium and electronic equipment

The invention is a divisional application with the application date of 2020, 06 and 28, and the application number of 202010598972.0, and is named as a load balancing control method and device, a storage medium and an electronic device.

Technical Field

The invention relates to the field of computers, in particular to a load balancing control method and device, a storage medium and electronic equipment.

Background

With the development of information technology, the total capacity of information storage has increased explosively. PB-level data-intensive applications like cloud computing generate a large number of requests for backend storage. The Object Storage system realizes physical separation of MetaData and file contents (data), the MetaData is managed by a MetaData Server (MDS for short), the data is managed by an Object Storage Device (OSD for short), the Object Storage system has quick access to a Storage Area Network (SAN for short) and secure data sharing of Network Attached Storage (NAS for short), and the Object Storage system is the future of a big data Storage technology.

In a distributed file system, the object storage system IO performance is largely dependent on the metadata processing performance. However, file systems that employ centralized metadata processing are limited in their performance and availability to a single MDS. Therefore, a distributed processing scenario, represented by ceph, in which multiple servers form an MDS cluster is receiving more and more attention. The problem of load balancing on the MDS is particularly important in order to maximize the processing performance of the MDS cluster metadata and avoid the MDS becoming a system bottleneck under certain conditions.

At present, in the load balancing method provided by the related art, a single MDS is not really overloaded, and blind activation of load migration not only causes waste of the capacity of the MDS, but also unnecessary load migration consumes resources such as originally tense network bandwidth, Central Processing Unit (CPU), memory, and the like.

In view of the above problems, no effective solution has been proposed.

Disclosure of Invention

The embodiment of the invention provides a load balancing control method and device, a storage medium and electronic equipment, and aims to at least solve the technical problem that object storage resources are wasted in the load balancing control method provided by the related art.

According to an aspect of the embodiments of the present invention, there is provided a load balancing control method, including: acquiring an average load value of load values of each metadata server in a metadata server cluster and a current load value of a current metadata server in the metadata server cluster; determining that the current metadata server is in a relative overload state if the current load value of the current metadata server is greater than the average load value; comparing the current load value with a load upper limit value configured for the current metadata server under the condition that the current metadata server is in a relative overload state; determining that the current metadata server is in an overload state when the current load value is greater than the load upper limit value configured for the current metadata server; and migrating the metadata in the current metadata server in the overload state.

According to an aspect of the embodiments of the present invention, there is provided a load balancing control method, including: acquiring load values of each metadata server in a metadata server cluster; determining that the current metadata server in the metadata server cluster is in an overload state according to the load value; determining a target metadata sub-tree to be migrated from the current metadata server, and determining a target metadata server from other metadata servers in the metadata server cluster except the current metadata server, wherein the target metadata server is in a non-overload state: and migrating the target metadata subtree to the target metadata server.

According to another aspect of the embodiments of the present invention, there is also provided a load balancing control apparatus, including: the metadata server cluster comprises an acquisition unit and a processing unit, wherein the acquisition unit is used for acquiring the average load value of the load values of each metadata server in the metadata server cluster and the current load value of the current metadata server in the metadata server cluster; a first determining unit, configured to determine that the current metadata server is in a relative overload state when the current load value of the current metadata server is greater than the average load value; a comparing unit, configured to compare the current load value with a load upper limit value configured for the current metadata server when the current metadata server is in a relative overload state; a second determining unit, configured to determine that the current metadata server is in an overload state when the current load value is greater than the load upper limit value configured for the current metadata server; and a migration unit, configured to perform metadata migration on the current metadata server in the overload state.

According to another aspect of the embodiments of the present invention, there is also provided a load balancing control apparatus, including: an obtaining unit, configured to obtain a load value of each metadata server in the metadata server cluster: a first determining unit, configured to determine, according to the load value, that a current metadata server in the metadata server cluster is in an overload state; a second determining unit, configured to determine a target metadata sub-tree to be migrated from the current metadata server, and determine a target metadata server from metadata servers other than the current metadata server in the metadata server cluster, where the target metadata server is in a non-overload state: and the migration unit is used for migrating the target metadata subtree to the target metadata server.

According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the load balancing control method when running.

According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor is configured to execute the load balancing control method through the computer program.

In the embodiment of the present invention, after the load state information of each metadata server in the metadata server cluster is obtained, an average load value of the metadata server cluster, a current load value of a current metadata server, and a load upper limit value configured for the current metadata server are determined. Then, in a case where it is determined that the current load value is greater than the average load value and greater than the load upper limit value, it is determined that migration is performed on the current metadata server. The method combines the relative overload evaluation absolute overload evaluation result with the absolute overload evaluation result, not only can more fully utilize the capability of the current metadata server, but also can better balance the distribution characteristic and the local characteristic of the load in the metadata server cluster, and finally comprehensively improves the overall performance of the metadata server cluster, thereby avoiding the problem of object storage resource waste in the load balancing control process.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:

fig. 1 is a schematic diagram of an application environment of an alternative load balancing control method according to an embodiment of the present invention;

FIG. 2 is a flow chart of an alternative load balancing control method according to an embodiment of the present invention;

FIG. 3 is a flow chart of another alternative load balancing control method according to an embodiment of the present invention;

FIG. 4 is a flow chart of yet another alternative load balancing control method according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of an alternative load balancing control method according to an embodiment of the present invention;

FIG. 6 is a reference schematic diagram of an alternative load balancing control method according to an embodiment of the invention;

FIG. 7 is a reference schematic diagram of another alternative load balancing control method according to an embodiment of the invention;

FIG. 8 is a migration diagram of an alternative load balancing control method according to an embodiment of the present invention;

FIG. 9 is a migration diagram of an alternative load balancing control method according to an embodiment of the present invention;

FIG. 10 is a flow chart of yet another alternative load balancing control method according to an embodiment of the present invention;

FIG. 11 is a flow chart of yet another alternative load balancing control method according to an embodiment of the present invention;

FIG. 12 is a schematic structural diagram of an alternative load balancing control apparatus according to an embodiment of the present invention;

fig. 13 is a schematic structural diagram of an alternative electronic device according to an embodiment of the invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

The embodiments provided in the present application can be used in, but are not limited to, a Cloud Storage (Cloud Storage) which is a new concept extended and developed from a Cloud computing concept, and a distributed Cloud Storage system (hereinafter referred to as a Storage system) refers to a Storage system which integrates a large number of Storage devices (Storage devices are also referred to as Storage nodes) of various types in a network through functions such as cluster application, grid technology, and distributed Storage file system, and provides a data Storage function and a service access function to the outside together through application software or application interfaces.

At present, a storage method of a storage system is as follows: logical volumes are created, and when created, each logical volume is allocated physical storage space, which may be the disk composition of a certain storage device or of several storage devices. The client stores data on a certain logical volume, that is, the data is stored on a file system, the file system divides the data into a plurality of parts, each part is an object, the object not only contains the data but also contains additional information such as data Identification (ID), the file system writes each object into a physical storage space of the logical volume, and the file system records storage location information of each object, so that when the client requests to access the data, the file system can allow the client to access the data according to the storage location information of each object.

The process of allocating physical storage space for the logical volume by the storage system specifically includes: physical storage space is divided in advance into stripes according to a group of capacity measures of objects stored in a logical volume (the measures often have a large margin with respect to the capacity of the actual objects to be stored) and Redundant Array of Independent Disks (RAID), and one logical volume can be understood as one stripe, thereby allocating physical storage space to the logical volume.

According to an aspect of the embodiments of the present invention, a load balancing control method is provided, and optionally, as an optional implementation manner, the load balancing control method may be applied, but not limited, to the environment shown in fig. 1. A MetaData Server cluster (Multi-MDS) consisting of a plurality of MetaData servers (MDSs) may be included in this environment, but is not limited to, as shown in fig. 1, n MetaData servers MDS including M1 through Mn. Each metadata server MDS is a core component for managing and providing metadata services in the file storage system, is used for constructing a file system hierarchical structure on the storage system, is responsible for creating and deleting directories and files, quota management, access control and the like, ensures data consistency among the MDSs through a distributed locking mechanism, and commonly undertakes management and metadata service tasks of the file storage system through mutual cooperation among cluster members.

In the embodiment of the present application, each MDS in the metadata server cluster (Multi-MDS) M1 through Mn shown in fig. 1 is individually decided, and there is no master control node in the cluster. Furthermore, each node MDS in the MDS cluster exchanges the latest load condition with each other through heartbeat information, after the load conditions of all other MDSs are collected, overload evaluation is started to judge whether the node MDS is overloaded or not, and if the node MDS is overloaded, a proper metadata subtree is selected according to a certain strategy and is migrated to other underloaded MDSs.

In addition, in this embodiment, if the node MDS in the Multi-MDS exits, no configuration or operation is required; if a new node MDS is added, only a group of reasonable load operation parameters, such as a load capacity value, a migration load proportion Ratio and an upper limit of the number of the migration subtrees, need to be manually configured for the new MDS. In addition, in the present embodiment, after increasing the number of node MDSs of the MDS cluster, the cluster performance will also increase linearly with the number of node MDSs.

Optionally, in this embodiment, the MDF metadata server may be a cloud server that provides basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, web service, cloud communication, middleware service, domain name service, security service, content delivery web server CDN, and big data and artificial intelligence platform. But is not limited thereto. The server may be directly or indirectly connected through wired or wireless communication, and the application is not limited herein. The above is merely an example, and this is not limited in this embodiment.

Optionally, as an optional implementation manner, as shown in fig. 2, the load balancing control method includes:

s202, acquiring heartbeat information which is periodically sent by other metadata servers except the current metadata server in the metadata server cluster, wherein the heartbeat information carries load state information of the metadata server;

s204, determining the average load value of the metadata server cluster according to the respective load state information of other metadata servers and the current load state information of the current metadata server;

s206, determining that the current metadata server is in an overload state under the condition that the current load state information indicates that the current load value of the current metadata server is greater than the average load value and is greater than the load upper limit value configured for the current metadata server;

s208, determining a target metadata subtree to be migrated from the current metadata server, and determining a target metadata server from other metadata servers, wherein the load value of the target metadata server is smaller than the average load value;

s210, migrating the target metadata subtree to a target metadata server.

Optionally, in this embodiment, the load balancing control method may be applied to, but not limited to, a cephFS multi-mds dynamic load balancing process. The comparison result between the current load value of the current metadata server in the metadata server cluster and the average load value of the metadata server cluster and the comparison result between the current load value and the load upper limit value of the current metadata server are combined, so that the combination of the relative overload evaluation absolute overload evaluation result and the absolute overload evaluation result is realized, the cephFS multi-mds overload evaluation process is optimized, a relative overload evaluation method is not adopted any more, the problem of waste of object storage resources is avoided, and unnecessary consumption and occupation of resources such as network bandwidth, a CPU (Central processing Unit), an internal memory and the like are reduced.

It should be noted that, in order to maintain load balance of the MDS cluster in the related art, a relative overload evaluation is usually directly adopted in the MDS cluster, and the nodes MDS in the MDS cluster are simply divided into two MDS sets, i.e., an overload set and a light load set. However, the distribution characteristic of the metadata in the MDS cluster often sacrifices the local (locality) characteristic of the metadata. The overall capacity of the MDS cluster can be improved by enhancing the distribution characteristics of the metadata, but the overall capacity of the MDS cluster is reduced by destroying the local characteristics of the metadata, and the final overall capacity of the MDS cluster is determined by the metadata distribution and the local characteristics. Relative overload MDSs may not be truly overloaded if they are absolutely capable of bearing the load according to the MDS itself. In this case, no metadata subtree migration is necessary for the relatively overloaded MDS at all, because the metadata subtree migration can further improve the metadata distribution characteristics, but also degrade the metadata locality characteristics. At this time, the positive contribution of the improvement of the metadata distribution characteristics to the overall capacity of the MDS cluster is far less than the negative influence of the overall capacity of the MDS cluster caused by the reduction of the metadata local characteristics.

Thus, to overcome the above-mentioned problems, the overload assessment process provided in the embodiments of the present application may include, but is not limited to: and evaluating relative overload and absolute overload, wherein the relative overload is evaluated to judge whether the MDS is overloaded according to the MDS load value and the multi-MDS cluster average load value. The absolute overload evaluation is to evaluate whether the MDS is overloaded or not according to the MDS load value and the absolute load bearing capacity of the MDS.

Specifically, the relative overload evaluation is to use the average load value of the MDS cluster as a relative overload line, divide each member MDS in the MDS cluster into two sets, i.e., a relatively light load and a relatively overload, determine that the MDS with the load value exceeding the relative overload line is the relative overload, and determine that the MDS with the load value lower than the relative overload line is the relatively light load. Because the relatively overloaded MDS bears more load than the relatively lightly loaded MDS, the purpose of averagely bearing the load by the MDS cluster members needs to be achieved by migrating the load to the relatively lightly loaded MDS. The absolute overload evaluation refers to comparing the real-time state parameter of the MDS with the upper limit of the parameter set by the MDS to judge whether the MDS is overloaded, wherein the overload is determined when the real-time state parameter of the MDS exceeds the upper limit of the set parameter, and the light load is determined when the real-time state parameter of the MDS is lower than the upper limit of the set parameter.

In this embodiment, the manner of combining the relative overload evaluation absolute overload evaluation result with the absolute overload evaluation result is to perform the relative overload evaluation on the MDS first, and if the MDS is determined as the relative overload, the absolute overload evaluation is started, and only under the condition that the MDS is determined as the relative overload and the absolute overload at the same time, the MDS can be finally determined to be in the overload state. Load migration can be started further, and load migration is not started in other cases, namely, the load migration is considered to be in a light load state.

Optionally, in this embodiment, the load status information may include, but is not limited to, at least one of the following: an amount of load indicating a load capacity of the metadata server, a number of requests processed by the metadata server per unit time (as represented by a req _ rate), and a CPU usage of the metadata server. That is, the current load value of the current metadata server (which may be simply referred to as the current MDS) may be, but is not limited to: current load, current request amount in unit time, current CPU utilization rate and the like.

Further, in this embodiment, the state parameter used for comparing the current load value with the average load value in the relative overload evaluation process and the state parameter used for comparing the current load value with the upper limit value in the absolute overload evaluation process may be the same state parameter or different state parameters. For example, in the determination process of the relative overload, when the current load value is the current load amount of the current MDS, the average load value may be the average load amount of the MDF cluster; in the absolute overload determination process, when the current load value is the current load amount of the current MDS, the corresponding load upper limit value may be the load upper limit amount (i.e., the maximum load amount) allowed to be borne by the current MDF. For another example, in the determination process of the relative overload, when the current load value is the current load amount of the current MDS, the average load value may be the average load amount of the MDF cluster; in the absolute overload determination process, in the case that the current load value is the number of processing requests in the current MDS unit time (as represented by req _ rate), the corresponding load upper limit value may be the upper limit amount of processing requests in the current MDF unit time (i.e., the maximum request amount req _ rate _ max). The above is an example, and this is not limited in this embodiment.

Optionally, in this embodiment, the manner of determining the target metadata subtree to be migrated from the current metadata server may include, but is not limited to, one of the following policies: 1) determining according to a preset upper limit of the migration load amount; 2) the number of the migration subtrees is determined according to a preset upper limit of the number of the migration subtrees. Therefore, the problem that the clients are too long to be blocked when accessing the subtrees due to the fact that the subtrees in migration are locked and protected due to the fact that a large number of subtree load migrations are involved in a single load balancing process is avoided. And furthermore, the adjustable and controllable migration load capacity of single load balance is realized, and the effect of shortening the pause time of the client when accessing the subtrees is achieved.

The description is made with reference to the example shown in fig. 3:

in step S302, determine whether the timer (ticket) of the current metadata server MDS in the metadata server cluster (multi-MDS) expires? And executing the steps S304-S306 in the case of arrival, wherein the current MDS sends heartbeat information to other MDSs in the multi-MDSs and receives heartbeat information sent by the other MDSs. Then, in step S308, it is determined whether all heartbeat information sent by all MDSs in the metadata server cluster (multi-MDS) is received, and if not, the process returns to step S306; if all heartbeat information is received, step S310 is executed to perform relative overload evaluation and absolute overload evaluation on the current MDS. If the current MDS is determined not to be overloaded, the process jumps to step S320, and ends to wait for the next ticket to arrive at time. If the current MDS is determined to be in the overload state, executing steps S314-S320, performing migration target matching to determine a target MDS, selecting a migration subtree, then executing migration to migrate the selected migration subtree into the target MDS, ending after the migration is completed, and waiting for the arrival of the next ticket.

The content and sequence of the steps shown in fig. 3 are examples, and this embodiment is not limited thereto.

By the embodiment provided by the application, after the load state information of each metadata server in the metadata server cluster is obtained, the average load value of the metadata server cluster, the current load value of the current metadata server and the load upper limit value configured for the current metadata server are determined. Then, in a case where it is determined that the current load value is greater than the average load value and greater than the load upper limit value, it is determined that migration is performed on the current metadata server. The method combines the relative overload evaluation absolute overload evaluation result with the absolute overload evaluation result, not only can more fully utilize the capability of the current metadata server, but also can better balance the distribution characteristic and the local characteristic of the load in the metadata server cluster, and finally comprehensively improves the overall performance of the metadata server cluster, thereby avoiding the problem of object storage resource waste in the load balancing control process.

As an optional scheme, the determining that the current metadata server is in an overload state includes:

s1, analyzing the respective load state information of other metadata servers to obtain the object load value corresponding to each metadata server to obtain a plurality of object load values;

s2, obtaining the average value of the load values of the plurality of objects as the average load value;

s3, comparing the current load value with the average load value;

s4, comparing the current load value with the load upper limit value under the condition that the current load value is larger than the average load value;

s5, in case that the current load value is greater than the load upper limit value, determining that the current metadata server is in an overload state.

Optionally, in this embodiment, after comparing the current load value and the average load value, the method further includes: under the condition that the current load value is smaller than the average load value, determining that the current metadata server is in a first light load state; after comparing the current load value with the load upper limit value, the method further comprises the following steps: and under the condition that the current load value is smaller than the load upper limit value, determining that the current metadata server is in a second light load state.

It should be noted that the first light load state may refer to, but is not limited to, that the load level of the current MDS has not reached the average level of the metadata server cluster (multi-MDS), and is in an absolute light load state. The second light load state may refer to, but is not limited to, the current MDS reaching a load level that exceeds an average level of a metadata server cluster (multi-MDS), but has not reached its upper load limit level, and is in a light load state relative to itself. Storage resources in a metadata server cluster (multi-MDS) are maximally utilized by differentiating different light load states to facilitate migration of different metadata subtrees for the MDSs.

The description is made with reference to the example shown in fig. 4: assume a current metadata server in a metadata server cluster (multi-MDS), with MDS₀For example. When MDS₀After receiving other MDS heartbeat information of the MDS cluster, obtaining load state information of other MDSs by analyzing the heartbeat information, wherein the load state information comprises a load value meta _ load of each MDS, so that an average load value avg _ meta _ load of the MDS cluster can be calculated conveniently. The various steps shown in fig. 4 are then performed:

in step S402, the MDS₀Its own load value meta _ load₀Comparing with the average load value avg _ meta _ load, if meta _ load₀If the value is less than avg _ meta _ load, MDS₀If the light load is determined, the latest light load is updated in step S404-1. Otherwise determining MDS₀Is relatively overloaded and step S404-2 is performed. When MDS₀When it is relatively overloaded, MDS₀Load migration is not initiated blindly and immediately, but rather is subject to an absolute overload assessment. The absolute overload evaluation is assumed to be performed by taking the processing request amount req _ rate per unit time as an example. And MDS₀The upper limit of the processing request amount per unit time that has been set in advance is req _ rate _ max. The decision process may be as follows:

MDS₀periodically records the unit time processing request amount req _ rate₀And as step S404-2, the current MDS₀Req _ rate of₀Comparing with the req _ rate _ max, determining the req _ rate₀>In the case of req _ rate _ max, the MDS is determined₀If the overload is absolute overload, step S406 is executed, otherwise, the light load is executed, and the step S404-1 is skipped. If MDS₀Are judged as relative overload and absolute overload at the same time, then MDS₀And finally, the load is judged to be in an overload state, and then the load migration process can be started.

Further in this example, further statistics may be made on the number of times the above determination is in the overload state, such as assuming that the continuous overload threshold is set to 3 times. Then, in step S406, the current MDS is determined₀Has been continuously overloaded more than 3 times, the current MDS is determined in step S408₀And the overload state is achieved, otherwise, the light load state is determined as step S410. Finally, the overload evaluation is ended as step S412.

In this embodiment, MDS is used₀Determining the MDS only if both a relative overload and an absolute overload are determined₀In an overload state. Can be used in MDS₀In the case of an overload state in a single decision process, the overload evaluation is terminated, which may also be in the MDS₀Under the condition that the overload state is judged in the multiple judging processes, the overload state is judged againThe overload evaluation is ended. Therefore, the MDS in the overload state is filtered to a certain extent by using the overload threshold value, and the problem of resource waste caused by misjudgment is avoided.

Furthermore, in this embodiment, the MDS is₀For any MDS in the metadata server cluster (multi-MDS), the process for determining overload evaluation of other MDSs may refer to the flow shown in fig. 4, which is not described in detail in this embodiment.

It should be noted that, in order to avoid bottlenecks and failures caused by single-point MDSs, the cephFS MDS cluster is not provided with a global scheduler for specially managing and maintaining load distribution among the MDSs in the cluster, but is independently decided by each MDS. The Multi-MDS exchanges the latest load state information with each other by periodically sending heartbeat information to acquire the load value of each MDS, and starts overload evaluation after collecting heartbeat information of all other MDSs, and the overload MDS starts a load migration process.

In the load balancing control method provided by the related art, whether to migrate a metadata sub-tree is mainly determined based on the determination result of the relative overload evaluation. For example, if MDS₀Is judged as a relative overload, MDS₀The load migration process is directly started, relatively light-load MDS in the cluster is used as target MDS for load migration, and the difference between the load value of the MDS and the average load value of the MDS cluster is used as the total load migration amount. Then, MDS₀And selecting target metadata subtrees to be migrated to load out according to a certain strategy, and sequentially migrating the target metadata subtrees to the target MDS until the current load migration amount reaches the total load migration amount.

The relative overload evaluation here can be as shown in fig. 5(a), and each MDS in the MDS cluster is divided into two sets, i.e., a light load set and an overload set, by taking the average load value of the MDS cluster as a relative overload line. The MDS with the load value exceeding the relative overload line is a relative overload, the MDS with the load value lower than the relative overload line is a relative light load, and because the relative overload MDS bears more load than the relative light load MDS, the purpose that the MDS cluster members bear the load on average is achieved by migrating the load to the relative light load MDS.

The dynamic load balancing control method provided in the embodiment of the present application may be as shown in fig. 5(b), and combine two dimensions of relative overload and absolute overload, further subdivide the situation of 50% relative overload on the basis of the relative overload evaluation result, and finally determine that the MDS (located in quadrant 2) that is relatively overloaded but absolutely lightly loaded is light load, rather than overloaded. The MDS that is relatively overloaded and absolutely overloaded (located in quadrant 1) is finally determined to be in an overloaded state.

Through the embodiment that this application provided, combine relative overload and absolutely overload two dimensions to determine the real metadata server that is in overload state in the metadata server cluster more accurately to make every metadata server can obtain make full use of, when avoiding unnecessary wasting of resources, improve each metadata server's utilization ratio.

As an optional scheme, determining a target metadata sub-tree to be migrated from the current metadata server includes:

s1, acquiring a load difference value between the current load value and the average load value;

s2, determining the load difference as a reference migration load quantity to be migrated by the current metadata server;

s3, acquiring the migration parameter upper limit value configured for the current metadata server;

and S4, determining a target metadata sub-tree to be migrated according to the reference migration load and the migration parameter upper limit value.

Optionally, in this embodiment, the migration parameter upper limit value may include, but is not limited to: the upper limit proportion of the migration load and the upper limit number of the migration subtrees. That is, one of the following migration control strategies may be employed, but is not limited to: and (4) a migration load upper limit control strategy or a migration subtree number upper limit control strategy.

It should be noted that, in the related art, it is common that the total outgoing load of the overloaded MDS is equal to the self load minus the average load of the MDS cluster, i.e., meta _ load-avg _ meta _ load is equal to the total outgoing load. For example, assume that MDS is included in a cluster of MDSs₀、MDS₁、MDS₂Wherein the current elementThe data server still uses MDS₀For example. Further, assume MDS as shown in FIG. 6(a)₀In an overload state, and the other two MDSs (i.e., MDSs)₁、MDS₂) In a light load state. MDS₀To the other two MDS (i.e. MDS)₁、MDS₂) Migrating loads (i.e., metadata subtrees), such as MDS shown in FIG. 6(b)₀Load shifting to MDS₁、MDS₂In order to achieve the purpose that 3 MDSs averagely bear the MDS cluster load as shown in fig. 6 (c).

When the load balancing method provided by the related art is adopted, MDS₀A plurality of metadata subtrees can be selected through a certain strategy and are sequentially migrated to other two MDSs until the sum of the load capacity of the migrated metadata subtrees reaches the MDS₀Until the total emigration loading. Therefore, the process of the overload MDS load balancing is finished by reaching the total emigration load of the overload MDS. But the migrated metadata subtrees are locked and protected during the migration process, so that the cephFS client has the problem of too long kathon time when accessing the metadata subtrees. Thus, the migration method shown in fig. 7(a) to 7(b) is not adopted in the embodiment of the present application, that is, the MDS is not adopted any more₀The metadata subtree in (1) is migrated to the other two MDSs (i.e., MDSs) in turn₁、MDS₂) The results are shown in FIG. 7 (c).

In order to solve the technical problems in the related art, in the embodiment of the present application, a migration parameter upper limit value (e.g., a migration load upper limit or a migration sub-tree number upper limit) is set for each MDS, and a single load balancing migration load is flexibly controlled by combining a reference migration load to be migrated by a current metadata server and the migration parameter upper limit value, so as to avoid a problem that a pause time is too long during access.

According to the embodiment provided by the application, the target metadata subtree to be migrated is determined according to the reference migration load and the migration parameter upper limit value. Therefore, the single load balance migration capacity is flexibly controlled and adjusted, and the blockage problem caused by the overlarge single load migration capacity is avoided.

As an optional scheme, determining a target metadata sub-tree to be migrated according to a reference migration load and a migration parameter upper limit value includes:

s1, under the condition that the upper limit value of the migration parameter is the upper limit ratio of the migration load, obtaining the product of the reference migration load and the upper limit ratio of the migration load to obtain a target migration load;

and S2, determining a target metadata sub-tree according to the target migration load.

Optionally, in this embodiment, determining the target metadata subtree according to the target migration load amount includes: traversing the metadata subtrees stored in the current metadata server, and repeatedly executing the following operations until the migration load capacity after migration is completed reaches the target migration load capacity:

s21, determining the current migration load quantity of the current completed migration;

s22, when the current migration load does not reach the target migration load, obtaining a current metadata sub-tree from the current metadata server, where a load corresponding to the current metadata sub-tree is a current data load;

and S23, determining the current metadata sub-tree as a target metadata sub-tree, and updating the current migration load after the current metadata sub-tree is migrated, wherein the difference value between the current load before updating and the current load after updating is the current data load.

The description is made with reference to the example shown in fig. 8: assuming MDS clusters include MDS₀、MDS₁、MDS₂Wherein the current metadata server remains with MDS₀For example. Further, assume MDS as shown in FIG. 8(a)₀In an overload state (over the average load line, and over the absolute load line), and the other two MDSs (i.e., MDSs)₁、MDS₂) In a light load state. Further, it is assumed that the transition parameter upper limit value is a transition load amount upper limit Ratio, and the transition load amount upper limit Ratio is 1/3.

In MDS₀By self loading metaThe average load Amount avg _ meta _ load (i.e., the average load line) is subtracted from the load to obtain the reference migration load Amount (the Amount is represented by a height L1 as shown in fig. 8 (a)). And determining a target migration load Amount (Amount) 1/3 according to the migration control strategy. The following operations are then repeatedly performed:

after each migration of a metadata sub-tree, the MDS₀The current migration load Amount cur _ Amount at which the migration has been currently completed will be calculated. If the current migration load Amount cur _ Amount<And (4) continuing to acquire the metadata subtree for migration if the target migration load Amount Amount 1/3. And if the current migration load Amount cur _ Amount>And (4) terminating the load balancing process when the target migration load Amount Amount is 1/3, which indicates that the migration is completed. As shown in fig. 8(b), migration results: migrating the target migration load Amount 1/3 (black solid square in the figure) to MDS₁In (1).

That is, after each subtree load is migrated, it is determined whether the current migration load amount currently completing migration reaches the target migration load amount, and if the current migration load amount reaches the target migration load amount, the load balancing process is immediately ended.

According to the embodiment provided by the application, the migration load upper limit proportion is used as the migration load amount control evaluation condition, and evaluation judgment is carried out after the migration is completed every time, so that the migration load of a single load balancing process is reduced, and the flexibility of load balancing control is improved.

s1, taking the upper limit number of the migration subtrees as the number of the target migration subtrees under the condition that the upper limit value of the migration parameter is the upper limit number of the migration subtrees;

s1, determining the target metadata subtree according to the number of the target migration subtrees.

Optionally, in this embodiment, determining the target metadata subtree according to the number of target migration subtrees includes: traversing the metadata subtrees stored in the current metadata server, and repeatedly executing the following operations until the number of the migration subtrees which have completed migration reaches the target number of the migration subtrees:

s21, determining the number of the current migration subtrees which have completed the migration;

s22, under the condition that the number of the current migration subtrees does not reach the number of the target migration subtrees, obtaining the current metadata subtrees from the current metadata server;

s23, determining the current metadata sub-tree as the target metadata sub-tree, and after the current metadata sub-tree is migrated, reducing the number of the current migration sub-tree by one.

The description is made with reference to the example shown in fig. 9: assuming MDS clusters include MDS₀、MDS₁、MDS₂Wherein the current metadata server remains with MDS₀For example. Further, assume MDS as shown in FIG. 8(a)₀In an overload state (over the average load line, and over the absolute load line), and the other two MDSs (i.e., MDSs)₁、MDS₂) In a light load state. Further assume that the migration parameter upper limit value is the migration subtree upper limit number, and the migration subtree upper limit number Max is 4.

In MDS₀The reference migration load Amount is obtained by subtracting the average load Amount avg _ meta _ load (i.e., the average load line) from the self load Amount meta _ load (the Amount is represented by a height L1 as shown in fig. 9 (a)). And determining the number of target migration subtrees to be 4 according to the migration control strategy. The following operations are then repeatedly performed:

after each migration of a metadata sub-tree, the MDS₀The number of currently migrated subtrees cur _ count _ subtree that have been currently migrated is counted. If the current migration subtree number cur _ count _ subtree<And 4, continuing to acquire the metadata subtrees for migration if the number of the target migration subtrees is 4. And if the current migration subtree number cur _ count _ subtree>If the number of the target migration subtrees is 4, the load balancing process is terminated, and the migration is finished. As shown in fig. 9(b), migration results: migrating metadata subtree 1, metadata subtree 3, and metadata subtree 6 to MDS₁In (2), migrating the metadata subtree 2 to the MDS₂In (2), the number of the current migration subtrees has reached the target number of migration subtrees 4.

That is, after the load of each subtree is migrated, it is determined whether the number of currently migrated subtrees that have been migrated currently reaches the number of target migration subtrees, and if the number of currently migrated subtrees reaches the number of target migration subtrees, the load balancing process is immediately ended.

According to the embodiment provided by the application, the upper limit number of the migration subtrees is used as the migration load control evaluation condition, and evaluation judgment is carried out after migration is completed every time, so that the migration load of a single load balancing process is reduced, and the flexibility of load balancing control is improved.

As an optional scheme, after determining the average load value of the metadata server cluster, the method further includes:

s1, obtaining the current load value according to the current load state information, wherein the current load state information includes the parameter value of at least one of the following state parameters: the number of processing requests of a current metadata server in unit time, the processor utilization rate of the current metadata server and the load capacity of the current metadata server;

and S2, acquiring the load upper limit value matched with the current load value.

It should be noted that the above-mentioned status parameter for making the absolute overload determination may include, but is not limited to, at least one of the following: the number of processing requests per unit time of the current metadata server, the processor usage of the current metadata server, and the load of the current metadata server. That is, each status parameter can be used for absolute overload evaluation, and the corresponding upper limit value of the status parameter can be used as the judgment threshold condition of the absolute overload. The load value may be, but not limited to, a weighted sum result or a product of at least two state parameters, which is finally used for determining. This is not limited in this embodiment.

Fig. 10 is a flowchart illustrating the entire load balancing control in the embodiment of the present application. Assuming MDS clusters include MDS₀、MDS₁、MDS₂Wherein the current metadata server remains with MDS₀For example. Assuming MDS₀In an overload state (over the average load line, and over the absolute load line), and the other two MDSs (i.e., MDSs)₁、MDS₂) In a light load state. Determining the need for MDS₀Partial metadata subtree migration in (1)_ToMDS₁、MDS₂In (1). The migration process may be as follows:

in step S1002, MDS is judged₀Is the middle metadata sub-tree traversal complete? In MDS₀In the case where the middle metadata subtree is not traversed to the end, step S1004 is performed to determine whether the current metadata subtree T0 is idle? If it is determined that the previous metadata sub-tree T0 is not idle, the process returns to step S1002. If it is determined to be idle, step S1006 is executed to use the metadata sub-tree T0 as the target metadata migration sub-tree to be migrated, and migrate the idle metadata sub-tree T0. And executing step S1008 after the migration is completed, determining whether a target migration load amount (calculated according to the upper limit proportion of the migration load amount) or the number of target migration subtrees (the upper limit number of the migration subtrees) is reached according to the migration control policy, if not, returning to step S1002, and if so, ending the step S1018.

In MDS₀In the case that the traversal of the middle metadata sub-tree is finished, step S1010 is performed to determine whether the traversal of the lightly loaded MDS is finished? Assuming that the current light load MDS is MDS₁If the light-load MDS has not been traversed, in step S1012, the MDS is determined₀Is the middle metadata sub-tree traversal complete? If MDS₀If the traversal of the middle metadata subtree is not completed, in step S1014, the current underloaded MDS is determined₁Whether the total load in (1) has reached the standard (i.e. whether MDS has been reached)₁Upper limit of load amount)? If not, it represents MDS₁The migration can be continuously accepted, and in step S1016, it is determined whether the target migration load amount (calculated according to the migration load amount upper limit ratio) or the target migration sub-tree number (migration sub-tree upper limit number) is reached according to the migration control policy. If the standard is reached, MDS is represented₁Cannot continue to accept the migration, and returns to obtain the next lightly loaded MDS, e.g., MDS₂To migrate metadata subtrees that have not completed migration to the MDS₂Thereby making it convenientIn completing the MDS₀Reference is made to the total migration of the migration load amount. And finally ends after the migration is completed as by step S1018.

In the embodiment of the application, dynamic load balancing optimization adjustment is performed on the metadata server cluster, so that the capacity of each MDS in the MDS cluster can be more fully mined and utilized, the waste of the capacity of the MDS and unnecessary metadata subtree migration are avoided, the service capacity of the cephFS multi-MDS cluster is comprehensively improved, and the overall performance of the cephFS is improved. In addition, under the same load condition, the capacity of each MDS in the cephFS multi-MDS cluster can be increased by 10% in the embodiment of the application. From another perspective, the cephFS multi-MDS cluster originally can provide the service capacity of N MDS nodes at most, and after the embodiment of the present application is adopted, the service capacity of N × 10% MDS nodes can be provided more. If the service capacity of the N MDS nodes meets the current service requirement, 1-1/(1+ 10%) -9% of MDS nodes can be removed, and the cluster construction cost is well saved.

According to another aspect of the embodiments of the present invention, as shown in fig. 11, there is also provided a load balancing control method, including:

s1102, acquiring an average load value of load values of each metadata server in the metadata server cluster and a current load value of a current metadata server in the metadata server cluster;

s1104, determining that the current metadata server is in a relative overload state under the condition that the current load value of the current metadata server is larger than the average load value;

s1106, comparing the current load value with the load upper limit value configured for the current metadata server under the condition that the current metadata server is in a relative overload state;

s1108, determining that the current metadata server is in an overload state under the condition that the current load value is greater than the load upper limit value configured for the current metadata server;

s1110, performing metadata migration on the current metadata server in the overload state.

Optionally, in this embodiment, the obtaining an average load value of load values of each metadata server in the metadata server cluster includes: acquiring object load values corresponding to each metadata server to obtain a plurality of object load values; and acquiring the average value of the plurality of object load values as an average load value.

Optionally, in this embodiment, after obtaining an average load value of load values of the metadata servers in the metadata server cluster, the method further includes: under the condition that the current load value is smaller than or equal to the average load value, determining that the current metadata server is in a first light load state; after comparing the current load value with the load upper limit value configured for the current metadata server, the method further includes: and under the condition that the current load value is less than or equal to the upper load limit value, determining that the current metadata server is in a second light load state.

For specific embodiments, reference may be made to the above method embodiments, which are not described herein again.

According to another aspect of the embodiments of the present invention, there is also provided a load balancing control method, including:

s1, acquiring the load value of each metadata server in the metadata server cluster;

s2, determining that the current metadata server in the metadata server cluster is in an overload state according to the load value;

s3, determining a target metadata sub-tree to be migrated from the current metadata server, and determining a target metadata server from other metadata servers except the current metadata server in the metadata server cluster, wherein the target metadata server is in a non-overload state:

s4, migrating the target metadata sub-tree to the target metadata server.

Optionally, in this embodiment, determining that a current metadata server in the metadata server cluster is in an overload state according to the load value includes: acquiring an average load value of load values of each metadata server in a metadata server cluster and a current load value of a current metadata server; and under the condition that the current load value of the current metadata server is greater than the average load value and is greater than the load upper limit value configured for the current metadata server, determining that the current metadata server is in an overload state.

Optionally, in this embodiment, determining the target metadata server from other metadata servers in the metadata server cluster besides the current metadata server includes: acquiring an average load value of load values of each metadata server in a metadata server cluster; determining a metadata server with a load value smaller than the average load value in the metadata server cluster as being in a first light load state; determining the metadata server with the load value smaller than the average load value and smaller than the correspondingly configured load upper limit value in the metadata server cluster as being in a second light load state; and determining a target metadata server from the metadata server in the first light load state and the metadata server in the second light load state.

Optionally, in this embodiment, determining the target metadata subtree to be migrated from the current metadata server includes:

traversing the metadata subtrees in the current metadata server in the overload state, and sequentially executing the following operations:

under the condition that the metadata subtrees in the current metadata server do not complete all traversal, obtaining the current metadata subtrees;

under the condition that the current metadata subtree is idle, determining the current metadata subtree as a target metadata subtree, and updating a migration state parameter, wherein the migration state parameter comprises at least one of the following parameters: the migration load capacity migrated by the current metadata server and the number of determined target metadata subtrees;

and under the condition that the updated migration state parameters do not reach the migration ending conditions, acquiring a next metadata subtree as the current metadata subtree.

Optionally, in this embodiment, migrating the target metadata subtree to the target metadata server includes:

traversing the target metadata server, and sequentially executing the following operations:

under the condition that the target metadata server does not complete all traversal, acquiring a current reference metadata server, and traversing a metadata sub-tree in the current reference metadata server;

under the condition that a metadata sub-tree in a current reference metadata server does not complete all traversal, acquiring the current load total amount of the current reference metadata server;

under the condition that the total current load of the current reference metadata server does not reach the upper limit value of the load configured for the current reference metadata server, determining the current reference metadata server as a current candidate metadata server for receiving a target metadata subtree to be migrated, and acquiring a current migration state parameter;

under the condition that the current migration state parameter does not reach the migration ending condition, performing metadata migration on the target metadata subtree to the determined idle metadata subtree in the current candidate metadata server;

stopping under the condition that the current migration state parameter reaches a migration ending condition, and performing metadata migration on the determined idle metadata subtrees in the current candidate metadata server;

under the condition that a metadata sub-tree in the current reference metadata server completes all traversals, acquiring a next reference metadata server from other metadata servers as the current reference metadata server;

and prompting to finish metadata migration under the condition that the target metadata server finishes all traversals or reaches a migration finishing condition, wherein the migration finishing condition comprises at least one of the following conditions: and the migrated migration load reaches the target load, and the number of the migrated target metadata subtrees reaches the number of the target migration subtrees.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the invention. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required by the invention.

According to another aspect of the embodiment of the present invention, there is also provided a load balancing control apparatus for implementing the load balancing control method. As shown in fig. 12, the apparatus includes:

1) a first obtaining unit 1202, configured to obtain heartbeat information periodically sent by other metadata servers in the metadata server cluster, where the heartbeat information carries load state information of the metadata server;

2) a first determining unit 1204, configured to determine an average load value of the metadata server cluster according to respective load state information of other metadata servers and current load state information of a current metadata server;

3) a second determining unit 1206, configured to determine that the current metadata server is in an overload state if the current load state information indicates that the current load value of the current metadata server is greater than the average load value and is greater than an upper limit of a load configured for the current metadata server;

4) a third determining unit 1208, configured to determine a target metadata sub-tree to be migrated from a current metadata server, and determine a target metadata server from other metadata servers, where a load value of the target metadata server is smaller than an average load value;

5) a migration control unit 1210 configured to migrate the target metadata subtree to the target metadata server.

Optionally, in this embodiment, the second determining unit 1206 includes:

1) the analysis module is used for analyzing the respective load state information of other metadata servers to obtain an object load value corresponding to each metadata server to obtain a plurality of object load values;

2) the first acquisition module is used for acquiring the average value of the load values of the plurality of objects as an average load value;

3) the first comparison module is used for comparing the current load value with the average load value;

4) the second comparison module is used for comparing the current load value with the load upper limit value under the condition that the current load value is larger than the average load value;

5) and the first determining module is used for determining that the current metadata server is in an overload state under the condition that the current load value is greater than the load upper limit value.

Optionally, in this embodiment, the method further includes: after the current load value and the average load value are compared, under the condition that the current load value is smaller than the average load value, the current metadata server is determined to be in a first light load state.

Optionally, in this embodiment, the method further includes: after comparing the current load value with the load upper limit value, the method further comprises the following steps: and under the condition that the current load value is smaller than the load upper limit value, determining that the current metadata server is in a second light load state.

Optionally, in this embodiment, the third determining unit 1208 includes:

1) the second obtaining module is used for obtaining a load difference value between the current load value and the average load value;

2) the second determining module is used for determining the load difference value as a reference migration load amount to be migrated by the current metadata server;

3) the third acquisition module is used for acquiring the migration parameter upper limit value configured for the current metadata server;

4) and the third determining module is used for determining a target metadata sub-tree to be migrated according to the reference migration load and the migration parameter upper limit value.

Optionally, in this embodiment, the third determining module determines the target metadata sub-tree to be migrated according to the reference migration load and the migration parameter upper limit value by: under the condition that the upper limit value of the migration parameter is the upper limit proportion of the migration load, obtaining the product of the reference migration load and the upper limit proportion of the migration load to obtain a target migration load; and determining a target metadata sub-tree according to the target migration load.

s1, determining the current migration load quantity of the current completed migration;

s2, when the current migration load does not reach the target migration load, obtaining a current metadata sub-tree from the current metadata server, where a load corresponding to the current metadata sub-tree is a current data load;

and S3, determining the current metadata sub-tree as a target metadata sub-tree, and updating the current migration load after the current metadata sub-tree is migrated, wherein the difference value between the current load before updating and the current load after updating is the current data load.

Optionally, in this embodiment, the third determining module determines the target metadata sub-tree to be migrated according to the reference migration load and the migration parameter upper limit value by: taking the upper limit number of the migration subtrees as the number of target migration subtrees under the condition that the upper limit value of the migration parameter is the upper limit number of the migration subtrees; and determining the target metadata subtrees according to the number of the target migration subtrees.

s1, determining the number of the current migration subtrees which have completed the migration;

s2, under the condition that the number of the current migration subtrees does not reach the number of the target migration subtrees, obtaining the current metadata subtrees from the current metadata server;

s3, determining the current metadata sub-tree as the target metadata sub-tree, and after the current metadata sub-tree is migrated, reducing the number of the current migration sub-tree by one.

Optionally, in this embodiment, the apparatus further includes: after the average load value of the metadata server cluster is determined, the current load value is obtained according to the current load state information, wherein the current load state information comprises parameter values of at least one of the following state parameters: the number of processing requests of the current metadata server in unit time, the processor utilization rate of the current metadata server and the load capacity of the current metadata server; and acquiring the load upper limit value matched with the current load value.

In this embodiment, reference may be made to the corresponding embodiment in the foregoing method for a specific embodiment, which is not described again in this embodiment.

According to another aspect of the embodiment of the present invention, there is also provided a load balancing control apparatus for implementing the load balancing control method. The device includes:

the metadata server cluster comprises an acquisition unit and a processing unit, wherein the acquisition unit is used for acquiring the average load value of the load values of all metadata servers in the metadata server cluster and the current load value of the current metadata server in the metadata server cluster;

the first determining unit is used for determining that the current metadata server is in a relative overload state under the condition that the current load value of the current metadata server is larger than the average load value;

the comparison unit is used for comparing the current load value with the load upper limit value configured for the current metadata server under the condition that the current metadata server is in a relative overload state;

the second determining unit is used for determining that the current metadata server is in an overload state under the condition that the current load value is larger than the load upper limit value configured for the current metadata server;

and the migration unit is used for performing metadata migration on the current metadata server in the overload state.

the acquisition unit is used for acquiring the load value of each metadata server in the metadata server cluster;

the first determining unit is used for determining that the current metadata server in the metadata server cluster is in an overload state according to the load value;

a second determining unit, configured to determine a target metadata sub-tree to be migrated from a current metadata server, and determine a target metadata server from other metadata servers in the metadata server cluster except the current metadata server, where the target metadata server is in a non-overload state;

and the migration unit is used for migrating the target metadata subtree to the target metadata server.

According to another aspect of the embodiment of the present invention, there is also provided an electronic device for implementing the load balancing control method, where the electronic device may be a server shown in fig. 1. As shown in fig. 13, the electronic device comprises a memory 1302 and a processor 1304, wherein the memory 1302 stores a computer program, and the processor 1304 is configured to perform the steps of any of the above method embodiments by the computer program.

Optionally, in this embodiment, the electronic device may be located in at least one network device of a plurality of network devices of a computer network.

Optionally, in this embodiment, the processor may be configured to execute the following steps by a computer program:

s1, acquiring heartbeat information periodically sent by other metadata servers except the current metadata server in the metadata server cluster, wherein the heartbeat information carries load state information of the metadata server;

s2, determining the average load value of the metadata server cluster according to the respective load state information of other metadata servers and the current load state information of the current metadata server;

s3, determining that the current metadata server is in overload state when the current load state information indicates that the current load value of the current metadata server is larger than the average load value and larger than the load upper limit value configured for the current metadata server;

s4, determining a target metadata sub-tree to be migrated from the current metadata server, and determining a target metadata server from other metadata servers, wherein the load value of the target metadata server is smaller than the average load value;

s5, migrating the target metadata sub-tree to the target metadata server.

Optionally, in this embodiment, the processor may be further configured to execute, by the computer program, the following steps:

s1, obtaining the average load value of the load values of each metadata server in the metadata server cluster and the current load value of the current metadata server in the metadata server cluster;

s2, determining that the current metadata server is in a relative overload state under the condition that the current load value of the current metadata server is larger than the average load value;

s3, comparing the current load value with the load upper limit value configured for the current metadata server under the condition that the current metadata server is in a relative overload state;

s4, determining that the current metadata server is in overload state when the current load value is larger than the upper limit value of the load configured for the current metadata server;

s5, performing metadata migration on the current metadata server in the overload state.

s4, migrating the target metadata sub-tree to the target metadata server.

Alternatively, it can be understood by those skilled in the art that the structure shown in fig. 13 is only an illustration, and the electronic device may also be a terminal device such as a smart phone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a palmtop computer, a Mobile Internet Device (MID), a PAD, and the like. Fig. 13 is a diagram illustrating a structure of the electronic device. For example, the electronics may also include more or fewer components (e.g., network interfaces, etc.) than shown in FIG. 13, or have a different configuration than shown in FIG. 13.

The memory 1302 may be configured to store software programs and modules, such as program instructions/modules corresponding to the load balancing control method and apparatus in the embodiments of the present invention, and the processor 1304 executes various functional applications and data processing by running the software programs and modules stored in the memory 1302, that is, implementing the load balancing control method. The memory 1302 may include high speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 1302 may further include memory located remotely from the processor 1304, which may be connected to the terminal over a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof. The memory 1302 may be used for information such as load status information of each MDS, but is not limited thereto. As an example, as shown in fig. 13, the memory 1302 may include, but is not limited to, a first obtaining unit 1102, a second determining unit 1104, a second determining unit 1106, a third determining unit 1108, and a migration control unit 1110 in the load balancing control apparatus. In addition, the load balancing control apparatus may further include, but is not limited to, other module units in the load balancing control apparatus, which is not described in detail in this example.

Optionally, the transmitting device 1306 is used for receiving or sending data via a network. Examples of the network may include a wired network and a wireless network. In one example, the transmission device 1306 includes a Network adapter (NIC) that can be connected to a router via a Network cable and other Network devices to communicate with the internet or a local area Network. In one example, the transmitting device 1306 is a Radio Frequency (RF) module, which is used to communicate with the internet in a wireless manner.

In addition, the electronic device further includes: a display 1308, configured to display load status information of each MDS; and a connection bus 1310 for connecting the respective module parts in the above-described electronic apparatus.

In other embodiments, the terminal device or the server may be a node in a distributed system, where the distributed system may be a blockchain system, and the blockchain system may be a distributed system formed by connecting a plurality of nodes through a network communication. Nodes can form a Peer-To-Peer (P2P, Peer To Peer) network, and any type of computing device, such as a server, a terminal, and other electronic devices, can become a node in the blockchain system by joining the Peer-To-Peer network.

According to an aspect of the application, a computer program product or computer program is provided, comprising computer instructions, the computer instructions being stored in a computer readable storage medium. The computer instructions are read by a processor of the computer device from the computer-readable storage medium, and the computer instructions are executed by the processor to cause the computer device to perform the steps provided in the various alternative embodiments of the cloud storage aspect described above.

Alternatively, in the present embodiment, the above-mentioned computer-readable storage medium may be configured to store a computer program for executing the steps of:

s5, migrating the target metadata sub-tree to the target metadata server.

Optionally, in this embodiment, the computer-readable storage medium may be further configured to store a computer program for executing the following steps:

s4, migrating the target metadata sub-tree to the target metadata server.

Alternatively, in this embodiment, a person skilled in the art may understand that all or part of the steps in the methods of the foregoing embodiments may be implemented by a program instructing hardware associated with the terminal device, where the program may be stored in a computer-readable storage medium, and the storage medium may include: flash disks, Read-Only memories (ROMs), Random Access Memories (RAMs), magnetic or optical disks, and the like.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

The integrated unit in the above embodiments, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in the above computer-readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing one or more computer devices (which may be personal computers, servers, network devices, etc.) to execute all or part of the steps of the method according to the embodiments of the present invention.

In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in the present application, it should be understood that the disclosed client may be implemented in other manners. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one type of division of logical functions, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims

1. A method for load balancing control, comprising:

acquiring an average load value of load values of each metadata server in a metadata server cluster and a current load value of a current metadata server in the metadata server cluster;

determining that the current metadata server is in a relative overload state if the current load value of the current metadata server is greater than the average load value;

comparing the current load value with a load upper limit value configured for the current metadata server under the condition that the current metadata server is in a relative overload state;

determining that the current metadata server is in an overload state when the current load value is greater than the upper load limit value configured for the current metadata server;

and carrying out metadata migration on the current metadata server in the overload state.

2. The method of claim 1, wherein obtaining an average load value of the load values of the metadata servers in the metadata server cluster comprises:

obtaining object load values corresponding to the metadata servers to obtain a plurality of object load values;

and acquiring the average value of the plurality of object load values as the average load value.

3. The method of claim 2,

after the obtaining of the average load value of the load values of the metadata servers in the metadata server cluster, the method further includes: determining that the current metadata server is in a first light load state when the current load value is less than or equal to the average load value;

after the comparing the current load value with the load upper limit value configured for the current metadata server, further comprising: and under the condition that the current load value is less than or equal to the load upper limit value, determining that the current metadata server is in a second light load state.

4. A method for load balancing control, comprising:

acquiring load values of each metadata server in a metadata server cluster;

determining that the current metadata server in the metadata server cluster is in an overload state according to the load value;

determining a target metadata sub-tree to be migrated from the current metadata server, and determining a target metadata server from other metadata servers in the metadata server cluster except the current metadata server, wherein the target metadata server is in a non-overload state:

migrating the target metadata subtree to the target metadata server.

5. The method of claim 4, wherein said determining that a current metadata server in the metadata server cluster is in an overloaded state based on the load value comprises:

acquiring an average load value of load values of each metadata server in the metadata server cluster and a current load value of the current metadata server;

determining that the current metadata server is in the overload state if the current load value of the current metadata server is greater than the average load value and the current load value is greater than an upper load limit configured for the current metadata server.

6. The method of claim 4, wherein determining a target metadata server from among other metadata servers in the metadata server cluster other than the current metadata server comprises:

acquiring an average load value of load values of each metadata server in the metadata server cluster;

determining a metadata server with a load value smaller than the average load value in the metadata server cluster as being in a first light load state;

determining the metadata server with the load value larger than the average load value and smaller than the load upper limit value of the corresponding configuration in the metadata server cluster as a second light load state;

and determining the target metadata server from the metadata server in the first light load state and the metadata server in the second light load state.

7. The method of claim 4, wherein determining the target metadata subtree to be migrated from the current metadata server comprises:

traversing the metadata subtrees in the current metadata server in the overload state, and sequentially performing the following operations:

under the condition that the metadata subtree in the current metadata server does not complete all traversal, acquiring the current metadata subtree;

determining the current metadata subtree as the target metadata subtree and updating a migration status parameter if the current metadata subtree is idle, wherein the migration status parameter includes at least one of: the migration load capacity migrated by the current metadata server and the determined number of the target metadata subtrees;

and under the condition that the updated migration state parameters do not reach the migration ending conditions, acquiring a next metadata sub-tree as the current metadata sub-tree.

8. The method of claim 7, wherein migrating the target metadata sub-tree to the target metadata server comprises:

under the condition that the metadata subtrees in the current reference metadata server do not complete all traversal, acquiring the current load total amount of the current reference metadata server;

determining that the current reference metadata server is a current candidate metadata server for receiving the target metadata subtree to be migrated and acquiring the current migration state parameter when the current load total amount of the current reference metadata server does not reach the load upper limit value configured for the current reference metadata server;

stopping under the condition that the current migration state parameter reaches the migration ending condition, and performing metadata migration on the determined idle metadata subtree in the current candidate metadata server;

under the condition that the metadata subtree in the current reference metadata server completes all traversals, acquiring a next reference metadata server from other metadata servers as the current reference metadata server;

and prompting to finish metadata migration under the condition that the target metadata server finishes all traversals or reaches a migration finishing condition, wherein the migration finishing condition comprises at least one of the following conditions: and the migrated migration load amount reaches a target load amount, and the number of the migrated target metadata subtrees reaches the number of the target migration subtrees.

9. A load balancing control apparatus, comprising:

a first determining unit, configured to determine that the current metadata server is in a relative overload state if the current load value of the current metadata server is greater than the average load value;

a comparing unit, configured to compare the current load value with a load upper limit value configured for the current metadata server when the current metadata server is in a relative overload state;

a second determining unit, configured to determine that the current metadata server is in an overload state when the current load value is greater than the load upper limit value configured for the current metadata server;

10. A load balancing control apparatus, comprising:

a first determining unit, configured to determine, according to the load value, that a current metadata server in the metadata server cluster is in an overload state;

a second determining unit, configured to determine a target metadata sub-tree to be migrated from the current metadata server, and determine a target metadata server from other metadata servers in the metadata server cluster except the current metadata server, where the target metadata server is in a non-overload state;

11. A computer-readable storage medium comprising a stored program, wherein the program when executed performs the method of any of claims 1 to 3 or the method of any of claims 4 to 8.

12. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, and wherein the processor is arranged to perform the method of any of claims 1 to 3 or the method of any of claims 4 to 8 by means of the computer program.