WO2020248586A1

WO2020248586A1 - Method for load balancing of agricultural machinery big data platform server

Info

Publication number: WO2020248586A1
Application number: PCT/CN2019/130636
Authority: WO
Inventors: 苑严伟; 徐玲; 庞在溪; 姜含露; 冀福华
Original assignee: 中国农业机械化科学研究院
Priority date: 2019-06-11
Filing date: 2019-12-31
Publication date: 2020-12-17
Also published as: CN110278251A

Abstract

A method for load balancing of an agricultural machinery big data platform server, comprising the following steps: in a scenario containing multiple servers, when user devices use the servers, virtualising a single server into a plurality of virtual servers, and initialising the price of the plurality of virtual servers; by means of a proportional fair algorithm, implementing resource allocation, each server broadcasting the lowest priced server quotation amongst the plurality of virtual servers thereof; a user device measures the communication rate of the adjacent server and the network bandwidth to the adjacent server and, on the basis of the server quotation and the communication rate and/or the network bandwidth, chooses to bid on the virtual server with the highest profitability; and, by means of an auction algorithm and on the basis of the bids of the user devices, the server selects the user device with the highest bid to match with the virtual server, such that the user device uses the corresponding virtual server.

Description

Server load balancing method of agricultural machinery big data platform

Technical field

The invention relates to agricultural informatization and network and communication technology, in particular to an agricultural machinery big data platform server load balancing method based on an auction algorithm.

Background technique

With the rapid development of agricultural informatization and the widespread application of agricultural machinery testing equipment, the application of "Internet + agricultural machinery" smart agricultural machinery in the agricultural field has emerged. Smart agricultural machinery makes full use of new-generation information technologies such as the Internet of Things, big data, and cloud computing to deeply integrate with the entire industrial chain of agricultural machinery. More and more cooperatives join the team of smart agricultural machinery. With the accumulation of work, the positioning information, status information, operation information and other data of agricultural machinery will show massive characteristics, especially during a certain operation season. The platform is facing a huge amount of visits and data; at the same time, agricultural machinery operations are increasingly dependent on online systems. A large number of key services require the system to have sufficient online rates, high efficiency and stability. These make the load of the agricultural machinery big data platform server heavier and heavier, and the performance requirements of the server's hardware and software are getting higher and higher.

As the server is in use, uneven distribution of user equipment will inevitably occur. Every user wants to use the server in the area that is closest to itself or has a higher data transmission rate. Due to the user's preference, the load of some servers reaches a threshold, the quality of service is worrying, and new users cannot use nearby servers.

Refer to Figure 1, which is a flow chart of load balancing in the prior art. The existing method of dealing with server load imbalance is mainly to select a server as a load balancer in a server cluster, which is responsible for optimizing the distribution of access requests among server groups. The client's traffic will first reach the load balancer, and the load balancer will distribute the traffic to different servers through a certain scheduling algorithm to improve the server's response speed and overall performance. Since all requests in this server cluster are passed through this load balancer, once the load balancer fails or is attacked by hackers, all requests will fail, and there are security issues; secondly, the load balancer wastes its own server Resources, relatively speaking, will lose part of the benefits.

Summary of the invention

The technical problem to be solved by the present invention is to address the above-mentioned defects of the prior art, and provide a method for load balancing of agricultural machinery big data platform servers based on auction algorithms. Load balancing is considered during the matching process of user equipment and servers, thereby avoiding unbalanced load distribution. Reasonable phenomenon.

In order to achieve the above objective, the present invention provides a method for load balancing of agricultural machinery big data platform servers, which includes the following steps:

S100. In a scenario containing multiple servers, when the user equipment uses the server, virtualize a single server into multiple virtual servers, and initialize the prices of the multiple virtual servers;

S200. Perform resource allocation through a proportional fair algorithm, and each server broadcasts the lowest-priced server quotation among the multiple virtual servers;

S300. The user equipment measures the communication rate of the neighboring server and the network bandwidth to the neighboring server, and selects the virtual server with the highest revenue according to the server offer and the communication rate and/or network bandwidth bid;

S400. Using an auction algorithm, the server selects the user equipment with the highest bid to match the virtual server according to the bid of the user equipment, so that the user equipment uses the corresponding virtual server.

In the above agricultural machinery big data platform server load balancing method, the server is virtualized into a corresponding number of the multiple virtual servers according to the number of the user equipment in the area where the server is located, and the server service If the number of user equipments in the range is n, then a single server is divided into n virtual servers, and the pricing strategy of each virtual server is as follows:

If the operating and idle costs of each virtual server are C _R and C _I , and C _R > C _I , the guaranteed reserve price of the server V _min = C _R- C _I ; for the virtual server j The kth said virtual server

Pricing

for:

Set the communication rate between the user equipment i and the server j as r _ij ; the server offers

The revenue expression of the user equipment for the server is defined as:

Among them, c is a set constant, and the value of c satisfies m _ij >0;

Set the bid of the user equipment i as

among them,

Represents the highest revenue of the user equipment i, and _m'i represents the second highest revenue of the user equipment i.

The above-mentioned agricultural machinery big data platform server load balancing method further includes the following steps:

S500. At the end of the single-round auction, it is judged whether the current round of auction results in a change in the topology of the connection. If the topology changes, the next round of auction is activated; if the topology remains unchanged, the algorithm converges and the auction is stopped.

In order to better achieve the above objectives, the present invention also provides a method for load balancing of agricultural machinery big data platform servers, in which the connection between the user equipment and the server is converted into a bipartite graph matching problem for processing, and the auction algorithm is combined, including the following step:

S10. Virtualizing one of the servers into multiple virtual servers, and setting an initial quotation for each of the virtual servers;

S20. The user equipment calculates the communication rate from itself to the neighboring server.

S30. When a single auction starts, each of the servers broadcasts the offer of the server with the lowest price among all the virtual servers;

S40. The user equipment calculates the revenue corresponding to each server according to the server quotation, and selects the two servers with the highest revenue and the second highest revenue from them;

S50. The user equipment uses the difference between the highest yield and the second highest yield of the server quotation as a value, and makes a bid to the corresponding server;

S60. The server selects the user equipment with the highest bid, allocates the virtual server with the lowest server bid among all virtual servers of the server to the user equipment, and raises the price of the virtual server to the price of the user equipment. Value

S70. The server broadcasts the final user equipment allocation result, and allocates resources to the user equipment using the server, and the user equipment uses the corresponding server according to the end user equipment allocation result.

In the above agricultural machinery big data platform server load balancing method, the server is virtualized into a corresponding number of virtual servers according to the number of user equipment in the area where the server is located, and the number of user equipment in the service range of the server is assumed Is n, then a single said server is correspondingly virtualized into n said virtual servers, and the pricing strategy for each said virtual server is as follows:

Assuming that the operating and idle costs of each virtual server are C _R and C _I , C _R > C _I , the reserve price of the server is V _min = C _R- C _I ; for the virtualized server Pricing of the k-th said virtual server

for:

The revenue expression of the user equipment for the server is defined as:

Among them, c is a set constant, and the value of c satisfies m _ij >0;

Set the bid of the user equipment i as

among them,

Represents its highest return, _m'i represents the next highest return.

In the above agricultural machinery big data platform server load balancing method, the server allocates resources to the user equipment using the server according to a proportional distribution fair algorithm.

In the above agricultural machinery big data platform server load balancing method, in the auction process, the user equipment uses the virtual server for non-permanent use, and the pricing of the virtual server increases according to the highest price of the user equipment.

In the above agricultural machinery big data platform server load balancing method, the server includes a server cluster, a physical server and/or a cloud server.

In the above-mentioned agricultural machinery big data platform server load balancing method, the load balancing process of the server does not involve the use of load balancing equipment.

S80. At the end of the single round of auction, determine whether the current round of auction results in a change in the connection topology. If the topology changes, perform step S30 to activate the next round of auction; if the topology remains unchanged, the algorithm converges and stops the auction.

The technical effects of the present invention are:

The invention has a wide range of application scenarios. Through the auction method, the virtual server is auctioned. During the auction process, the dependence of the user equipment on the target server is fully considered, and load balancing is realized in a distributed manner. Each step of allocating servers to user equipment is reasonably distributed, which achieves load balancing of servers to a large extent, greatly improves server resource utilization, and achieves proportional fairness among all user equipment. Taking a distributed approach to solve the load balancing problem reduces security risks and improves the processing speed of load balancing problems compared to using a load balancer. It meets the needs of more users who want to use closer servers, and provides higher-quality communication services. At the same time, the server's quotation and temporary connection topology are provided through the broadcast mechanism, which can greatly reduce the resulting communication overhead.

The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments, but not as a limitation to the present invention.

Description of the drawings

Figure 1 is a flow chart of load balancing in the prior art;

Figure 2 is a working principle diagram of an embodiment of the present invention;

Figure 3 is a schematic diagram of an application scenario of an embodiment of the present invention;

FIG. 4 is a schematic diagram of load balancing implemented in the process of a user equipment applying to use a server according to an embodiment of the present invention.

Wherein, the reference number

S1, S2 server

U1, U2, U3, U4 user equipment

r ₁₁ , r ₂₁ , r ₃₁ , r ₃₂ , r ₄₂ communication rate

Detailed ways

The structural principle and working principle of the present invention will be described in detail below with reference to the accompanying drawings:

Refer to Figure 2, which is a working principle diagram of an embodiment of the present invention. The load balancing method of the agricultural machinery big data platform server of the present invention may include the following steps:

Step S100: In a scenario containing multiple servers, when the user equipment uses the server, virtualize a single server into multiple virtual servers, and initialize the prices of the multiple virtual servers; wherein the server may Including server clusters, physical servers, and/or cloud servers, etc., and the load balancing process of the servers does not involve the use of load balancing equipment;

Step S200: Perform resource allocation through a proportional fairness algorithm, and each server broadcasts the lowest priced server quotation among the multiple virtual servers.

Step S300: The user equipment measures the communication rate of the neighboring server and the network bandwidth to the neighboring server, and selects the virtual server with the highest revenue according to the server offer and the communication rate and/or network bandwidth. Server bid

Step S400: Through an auction algorithm, the server selects the user equipment with the highest bid to match the virtual server according to the bid of the user equipment, so that the user equipment uses the corresponding virtual server.

It also includes the following steps:

Step S500: When the single round of auction ends, it is judged whether the current round of auction results in a change in the topology of the connection. If the topology changes, the next round of auction is activated; if the topology remains unchanged, the algorithm converges and the auction is stopped.

In another embodiment of the present invention, the agricultural machinery big data platform server load balancing method can convert the connection between the user equipment and the server into a bipartite graph matching problem for processing. At the same time, combined with the auction algorithm, the server distributes the fairness algorithm in proportion to The user equipment that uses the server allocates resources, where the server may include a server cluster, a physical server, and/or a cloud server, etc. The load balancing process of the server does not involve the use of load balancing equipment; specifically, it may include The following steps:

Step S10, virtualizing one of the servers into multiple virtual servers, and setting an initial quotation for each of the virtual servers;

Step S20: The user equipment measures the communication rate from itself to the surrounding neighboring servers;

Step S30: When a single auction starts, each server broadcasts the offer of the server with the lowest price among all the virtual servers;

Step S40: The user equipment calculates the revenue corresponding to each server according to the server quotation, and selects the two servers with the highest revenue and the second highest revenue from them;

Step S50: The user equipment uses the difference between the highest yield and the second highest yield of the server quotation as the value, and makes a bid to the corresponding server;

Step S60: The server selects the user equipment with the highest bid, allocates the virtual server with the lowest server bid among all virtual servers of the server to the user equipment, and increases the price of the virtual server to the user equipment The value of

Step S70: The server broadcasts the final user equipment allocation result, and allocates resources to the user equipment using the server, and the user equipment uses the corresponding server according to the end user equipment allocation result.

Wherein, the server can be virtualized into a corresponding number of virtual servers according to the number of user equipment in the area where the server is located, and assuming that the number of user equipment within the service range of the server is n, then a single server corresponds to switching. Divided into n virtual servers, the pricing strategy of each virtual server is as follows:

Assuming that the operating and idle costs of each virtual server are respectively C _R and C _I , C _R > C _I , the guaranteed reserve price of the server V _min = C _R- C _I ; for the virtual server j Pricing of the kth said virtual server

for:

The revenue expression of the user equipment for the server is defined as:

Among them, c is a set constant, and the value of c satisfies m _ij >0;

Set the bid of the user equipment i as

among them,

Represents its highest return, _m'i represents the next highest return.

Wherein, during the auction process, the use of the virtual server by the user equipment is non-permanent use, and the pricing of the virtual server is increased according to the highest price of the user equipment.

In this embodiment, the following steps may also be included:

Step S80: At the end of the single round of auction, judge whether the current round of auction results in a change in the connection topology. If the topology changes, the next round of auction will be activated; if the topology remains unchanged, the algorithm will converge, stop the auction, and the server will broadcast The end user equipment allocates the result and allocates resources for the user equipment using this server.

The working process of the present invention is explained in a specific embodiment as follows:

Step S10: In the initial stage of the auction, each server S is virtually divided into several virtual servers VS, and a single server S is to be divided into n virtual servers VS, where n is the number of user equipment U in the server service area. Set the initial quotation of the kth virtual server segmented by this server j

And V _min = C _R- C _I ; where V _min is the guaranteed base price of this server; C _R and C _I represent the cost of the virtual server when it is running and idle, C _R > C _I ;

Step S20: In the initial stage of the auction, each user equipment measures its communication rate to surrounding neighboring servers, and sets the communication rate between user equipment i and server j as r _ij ;

Step S30. When a single auction starts, each server broadcasts the lowest price among the virtual servers it owns. Taking server j as an example, then

Step S40, each user calculates the revenue corresponding to each server according to the quotation of the server in the area, and selects the two servers with the highest revenue and the second highest revenue from them. Taking user equipment i as an example, the user calculates the revenue corresponding to the jth server Quotation, calculate income

Among them, c is a set constant, and the value of c satisfies m _ij >0; user i selects the profit that makes the highest profit

And the second-highest return m′ _i .

Step S50: The user uses the difference between the highest income and the second highest income of his own equipment obtained in step S40 as his bid, and makes a bid to the corresponding server, which is

Step S60: The server selects the user equipment with the highest bid, allocates the virtual server with the lowest price to the user equipment, and raises the price of the virtual server to the bid of the user equipment that bids the server;

Step S70: The server sends the distribution result to the user equipment, and the single-round auction ends.

Step S80: When the single round of auction ends, it is judged whether the current round of auction results in a change in the connection topology. When a new connection topology is formed, a new round of auction starts, and the process goes to step S30; when the connection topology does not change, the auction ends, and the server broadcasts the end user equipment allocation result and allocates resources for the user equipment using the server.

The user equipment uses the corresponding server service according to the end user allocation result, and the server allocates bandwidth to the user equipment according to a proportional fair algorithm. Specifically, if there are n user equipments using the server j, the server j allocates bandwidth to the user in proportion to the communication rate of the user. For example, when the communication rates of users A and B are 2 Mbps and 5 Mbps, respectively, if the total bandwidth of the server is 3.5 Mbps, user A is allocated 1 Mbps and user B is allocated 2.5 Mbps of bandwidth.

In the above distribution and use process of the present invention, the use and load balance of each server do not need to be managed and controlled by the load balancer, but a distributed strategy is adopted. Each server does not execute the load balancing strategy when the threshold is reached, but takes the global load balancing problem into consideration when each user device requests it.

3 and FIG. 4, FIG. 3 is a schematic diagram of an application scenario of an embodiment of the present invention, and FIG. 4 is a schematic diagram of load balancing implemented in a process of a user equipment applying to use a server according to an embodiment of the present invention. In this embodiment, S1 and S2 are servers, and U1, U2, U3, and U4 are user equipment. Assume that the communication rates of each user equipment to the server are r ₁₁ =r ₃₁ =3, r ₂₁ =r ₃₂ =r ₄₂ =2, and c=3. Assume that the minimum price of all servers is V _min =1.

In this case, the specific implementation steps of load balancing are as follows:

1) In the initial state, the server S1 has 3 user devices U1, U2, and U3 in the area where it is located, so it is virtualized into 3 VSs.

Means; there are two user equipment U2 and U3 in the area where server S2 is located, so virtualized into two VSs, respectively

Said. According to the formula

Calculate its offer, where

Is a virtual server

s price.

See Table 1 below. Table 1 is a detailed quotation of each virtual server in this embodiment:

Table 1 Detailed quotation of each virtual server

1) At the beginning of the first round of auction, servers S1 and S2 respectively select the virtual server VS with the lowest price as the quotation, that is, price ₁ =1, price ₂ =1. The servers S1 and S2 respectively notify the users of their quotations by broadcasting.

2) The user equipment U1 calculates its highest profit as m*=2+log3. Since the user equipment U1 does not have the second highest profit, it submits the bid 2+log3 to S1. Similarly, the user equipment U2 and U4 also submit bids to S1 and S2 respectively. For the user equipment U3, because it is within the service range of the servers S1 and S2 where it is located, the highest revenue is calculated separately

The second highest return m′ ₃ =2+log2, so the user equipment U3 submits the bid to the server S1 as log3-log2.

3) The server S1 selects the user equipment with the highest price from all quotations, and assigns the virtual server with the lowest quotation to the user equipment. This round of user equipment U1 and virtual server

Match, user device U4 and virtual server

match. At the same time these two virtual servers

The offer of is modified to the bid of the corresponding user device. The first round of auctions is over.

4) The second round of auction begins. The servers S1 and S2 respectively select the virtual server VS with the lowest price, use the price of the virtual server VS as its own quotation, and broadcast the quotation.

5) In the aforementioned process, each user device submits a bid. The user equipment U2 and U3 respectively submit bids to the server S1.

6) The server S1 obtains the bids of the two user equipment U2 and U3, and the bid of the user equipment U2 is higher than that of the user equipment U3, thus obtaining the virtual server

usage of. Modify the virtual server

s price. The second round of auction ends.

7) The third round of auction begins. The servers S1 and S2 respectively select the virtual server VS with the lowest price, use the price of the virtual server VS as its own quotation, and broadcast the quotation.

8) The user equipment U3 calculates its profit and submits a bid to the server S2.

9) The server S2 accepts the bidding of the user equipment U3, and the user equipment U3 obtains the bid with the virtual server

Match. Modify the virtual server

s price. The third round of auction ends.

10) The fourth round of auctions begins, and after one round of auctions, there is no change in topology. The auction is over.

In the above process, the detailed quotations of each virtual server at each stage of the auction can be seen in Table 1.

In the above processes 2) and 5), both the user equipment U3 submits a bid to S1. In process 8), the user equipment U3 turns to S2 to bid. The reason is that the user equipment accessed by the server S1 increases after two auctions, and the bid price of the server S1 in the third round of auction is higher than that of the server S2, so the user equipment U3 turns to S2, the low-priced server. From this, it can be seen that the present invention plays the role of load balancing during the user application phase.

After each auction is over, it is necessary to determine whether the topology has changed. If it changes, the next round of auctions needs to be performed. If there is no change, the algorithm converges and the auction stops.

The invention is an agricultural machinery big data platform server load balancing method based on an auction algorithm. In a scenario containing multiple servers, when a user equipment uses the server, a single server is virtualized into multiple virtual servers, and resources are allocated through a proportional fair algorithm; Through the auction method, the user equipment uses a certain server, and each server broadcasts the price of the virtual server with the lowest price. The user equipment bids according to the network bandwidth to the adjacent server, and the server selects the user with the highest bid according to the user’s bid. The device matches the virtual server.

The invention realizes the load balance of the server, greatly improves the utilization rate of server resources, and realizes proportional fairness in all user equipment. Using a distributed method to solve the load balancing problem, compared with the use of load balancing equipment, reduces security risks and improves the processing speed of load balancing problems. It meets the needs of more users who want to use servers that are closer and have better communication quality.

Of course, the present invention can also have various other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding All changes and deformations shall belong to the protection scope of the appended claims of the present invention.

Claims

An agricultural machinery big data platform server load balancing method is characterized in that it comprises the following steps:

S100. In a scenario containing multiple servers, when the user equipment uses the server, virtualize a single server into multiple virtual servers, and initialize the prices of the multiple virtual servers;

S200. Perform resource allocation through a proportional fair algorithm, and each server broadcasts the lowest-priced server quotation among the multiple virtual servers;

S300. The user equipment measures the communication rate of the neighboring server and the network bandwidth to the neighboring server, and selects the virtual server with the highest revenue according to the server offer and the communication rate and/or network bandwidth bid;

S400. Using an auction algorithm, the server selects the user equipment with the highest bid to match the virtual server according to the bid of the user equipment, so that the user equipment uses the corresponding virtual server.
The agricultural machinery big data platform server load balancing method according to claim 1, wherein the server is virtualized into a corresponding number of multiple virtual devices according to the number of the user equipment in the area where the server is located. Server, assuming that the number of user equipment within the service range of the server is n, then a single server is divided into n virtual servers, and the pricing strategy of each virtual server is as follows:

If the operating and idle costs of each virtual server are C R and C I , and C R > C I , the guaranteed reserve price of the server V min = C R- C I ; for the virtual server j The kth said virtual server
Pricing
for:

Set the communication rate between the user equipment i and the server j as r ij ; the server offers
The revenue expression of the user equipment for the server is defined as:

Among them, c is a set constant, and the value of c satisfies m ij >0;

Set the bid of the user equipment i as
among them,
Represents the highest revenue of the user equipment i, and m'i represents the second highest revenue of the user equipment i.
The agricultural machinery big data platform server load balancing method according to claim 1 or 2, characterized in that it further comprises the following steps:

S500. At the end of the single-round auction, it is judged whether the current round of auction results in a change in the topology of the connection. If the topology changes, the next round of auction is activated; if the topology remains unchanged, the algorithm converges and the auction is stopped.
An agricultural machinery big data platform server load balancing method is characterized in that the connection between the user equipment and the server is converted into a bipartite graph matching problem for processing and combined with an auction algorithm, which specifically includes the following steps:

S10. Virtualizing one of the servers into multiple virtual servers, and setting an initial quotation for each of the virtual servers;

S20. The user equipment calculates the communication rate from itself to the neighboring server.

S30. When a single auction starts, each of the servers broadcasts the offer of the server with the lowest price among all the virtual servers;

S40. The user equipment calculates the revenue corresponding to each server according to the server quotation, and selects the two servers with the highest revenue and the second highest revenue from them;

S50. The user equipment uses the difference between the highest yield and the second highest yield of the server quotation as a value, and makes a bid to the corresponding server;

S60. The server selects the user equipment with the highest bid, allocates the virtual server with the lowest server bid among all virtual servers of the server to the user equipment, and raises the price of the virtual server to the price of the user equipment. Value

S70. The server broadcasts the final user equipment allocation result, and allocates resources to the user equipment using the server, and the user equipment uses the corresponding server according to the end user equipment allocation result.
The agricultural machinery big data platform server load balancing method of claim 4, wherein the server is virtualized into a corresponding number of virtual servers according to the number of user devices in the area where the server is located, and the server service range is set If the number of user equipments in is n, then a single server is divided into n virtual servers, and the pricing strategy for each virtual server is as follows:

Assuming that the operating and idle costs of each virtual server are respectively C R and C I , C R > C I , the guaranteed reserve price of the server V min = C R- C I ; for the split of the server j Pricing of the k-th said virtual server
for:

Set the communication rate between the user equipment i and the server j as r ij ; the server offers

The revenue expression of the user equipment for the server is defined as:

Among them, c is a set constant, and the value of c satisfies m ij >0;

Set the bid of the user equipment i as
among them,
Represents the highest revenue of the user equipment i, and m'i represents the second highest revenue of the user equipment i.
The agricultural machinery big data platform server load balancing method according to claim 4 or 5, wherein the server allocates resources to the user equipment using the server according to a proportional distribution fair algorithm.
The agricultural machinery big data platform server load balancing method according to claim 6, wherein in the auction process, the user equipment uses the virtual server for non-permanent use, and the virtual server is priced according to the user equipment The highest price increase.
The agricultural machinery big data platform server load balancing method according to claim 1, 2, 4, or 5, wherein the server includes a server cluster, a physical server, and/or a cloud server.
The agricultural machinery big data platform server load balancing method according to claim 1, 2, 4, or 5, wherein the load balancing process of the server does not involve the use of load balancing equipment.
The method for server load balancing of agricultural machinery big data platform according to claim 4 or 5, further comprising the following steps:

S80. At the end of the single round of auction, determine whether the current round of auction results in a change in the connection topology. If the topology changes, perform step S30 to activate the next round of auction; if the topology remains unchanged, the algorithm converges and stops the auction.