CN107222540B - Negative feedback-based server cluster grouping scheduling method - Google Patents
Negative feedback-based server cluster grouping scheduling method Download PDFInfo
- Publication number
- CN107222540B CN107222540B CN201710416047.XA CN201710416047A CN107222540B CN 107222540 B CN107222540 B CN 107222540B CN 201710416047 A CN201710416047 A CN 201710416047A CN 107222540 B CN107222540 B CN 107222540B
- Authority
- CN
- China
- Prior art keywords
- server
- period
- cpu utilization
- servers
- negative feedback
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/1008—Server selection for load balancing based on parameters of servers, e.g. available memory or workload
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1031—Controlling of the operation of servers by a load balancer, e.g. adding or removing servers that serve requests
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
Abstract
The invention discloses a server cluster grouping scheduling method based on negative feedback. The method comprises the following steps: 1) calculating the optimal server operation number of the kth period according to the historical operation state of the kth period of the server; 2) and obtaining the number of the servers to be started in the k +1 th period in a negative feedback mode according to the optimal server operation number and the operation state of the server in the k-th period. According to the invention, the resource utilization rate and the energy efficiency of the server cluster are improved by counting the historical operating conditions and automatically adjusting the number of the started servers in a negative feedback manner.
Description
Technical Field
The invention relates to a scheduling method of a server cluster, in particular to a server cluster method and system based on negative feedback.
Background
With the rapid development of network technologies such as cloud computing and internet of things, a server cluster technology is developed.
The server cluster is generally a cluster system in which a plurality of servers are connected through a high-speed network, and has the characteristics of high performance, high availability and high cost performance, so that the server cluster is widely applied. However, with the continuous increase of the scale of the data center, the energy consumption of the server occupies a large amount of enterprise investment, and energy-saving scheduling of a server cluster has become a problem of wide attention in the industry at present.
However, the server cluster also faces the problem of dynamic resource management under the condition of load intensity change, and if there is no cluster scheduling method based on grouping, the problems of resource utilization and performance reduction caused by too frequent switching of the start-stop states of the servers and unbalanced load may occur.
In the server cluster grouping scheduling, the number of servers to be started is an important parameter, the purpose of improving the utilization rate of the servers cannot be achieved due to the fact that the number of the started servers is too large, and the system requirements cannot be met due to the fact that the number of the started servers is too low.
Disclosure of Invention
In view of the above problems, the present invention provides a server cluster group scheduling method based on negative feedback, which improves resource utilization and energy efficiency of a server cluster by counting historical operating conditions and automatically adjusting the number of servers that are started in a negative feedback manner.
The technical scheme of the invention is as follows:
a server cluster grouping scheduling method based on negative feedback comprises the following steps:
1) calculating the optimal server operation number of the kth period according to the historical operation state of the kth period of the server;
2) and obtaining the number of the servers to be started in the k +1 th period in a negative feedback mode according to the optimal server operation number and the operation state of the server in the k-th period.
Further, the method for calculating the running number of the optimal servers comprises the following steps: setting the CPU utilization rate and the processing performance of the server acquired in the kth period as c ═ c respectively1,c2,…,ci,…,cn],p=[p1,p2,…,pi,…,pn]Wherein n is the total number of servers operating in the k period, ciFor the ith server CPU utilization, piFor the processing performance of the ith server, ci∈(0,1),p i0 or 1; according to piValue taking, namely dividing the set c into two sets c respectively0And c1I.e. the handling properties piAdding server CPU utilization with value 0 to set c0To process the property piAdding server CPU utilization with value 1 to set c1(ii) a Then dividing the CPU utilization rate into a plurality of intervals according to co=argmaxi[Ni 0/Ni 1]H is calculated to obtain the optimal operation parameter c of the k periodo(ii) a Wherein N isi 0Is a set c0The number of samples of the middle CPU utilization rate in the ith interval, Ni 1Is a set c1And the utilization rate of the middle CPU is the number of samples in the ith interval, and h is an interval coefficient.
Further, the processing performance is a packet loss rate, and is set to be 0 when there is a packet loss, and is set to be 1 when there is no packet loss.
Further, the interval coefficient h takes a value of 0.1.
Further, the method for obtaining the number of servers to be started in the k +1 th period in a negative feedback manner includes: and calculating the difference between the average CPU utilization rate of the server in the k period and the optimal CPU utilization rate in the k period, if the difference is greater than 0, increasing the number of the started servers in the k +1 period, and otherwise, reducing the number of the started servers.
Further, using the formulaCalculating the number M (k +1) of servers to be started in the k +1 th period; wherein M (k) is the number of servers that have been turned on in the k-th period,is the average CPU utilization in the k-th cycle, coIs the optimal operating parameter for the k-th cycle and α is the adjustment step in negative feedback.
The invention has two key technologies:
1) counting the optimal parameters of the server operation according to the historical operation state of the server;
2) and obtaining the number of the servers to be started in a negative feedback mode according to the optimal parameters of the server operation and the current operation state.
In view of the above, the main contents of the present invention are as follows:
acquiring the running state of the server in real time: and carrying out real-time statistics on the CPU utilization rate and the processing performance of the server at certain time intervals, wherein the processing performance comprises but is not limited to the network card packet loss rate. The processing performance and server running state information should be synchronized in time.
Counting the optimal operation parameters of the server:
setting the CPU utilization rate and the processing performance of the server acquired in a period of time as c ═ c1,c2,…,ci,…,cn],p=[p1,p2,…,pi,…,pn]Wherein c isi∈(0,1),p i0 or 1. According to the values of corresponding pi, c can be divided into two sets c0And c1. Meanwhile, the CPU utilization is divided into 10 intervals, and the number of samples in each interval is counted as follows.
The optimal operating parameters are:
co=argmaxi[Ni 0/Ni 1]h, h is the interval coefficient, the value in the invention is 0.1, and the coefficient ensures coIs between (0, 1).
Determining the current optimal server number in a negative feedback mode:
after each sampling is finished, calculating the difference between the current average CPU utilization rate and the optimal CPU utilization rate, if the current CPU utilization rate is higher than the optimal CPU utilization rate (the difference is more than 0), increasing the number of the started servers, otherwise, reducing the number of the servers. The specific calculation formula is as follows:
wherein M (k +1) is the number of servers that are turned on in the next operation cycle, M (k) is the number of servers that are currently turned on,is the average of the current CPU utilization, coIs the calculated optimum operating parameter and α is the step size.
Compared with the prior art, the invention has the following positive effects:
the invention can adaptively learn the optimal system operation parameters and dynamically schedule the system in the optimal working state, thereby having better performance power consumption ratio.
Drawings
Fig. 1 is a system block diagram of a negative feedback-based server cluster packet scheduling method in the present invention.
Fig. 2 is a schematic diagram of an optimal server number calculation algorithm for server grouping based on negative feedback in the present invention.
Detailed Description
The method of the present invention is described in detail below with reference to the accompanying drawings and examples.
As shown in fig. 1, the workflow of the method of the present invention in this embodiment is:
step 101: and in the system initialization stage, a load acquisition period is initialized, the adjustment step length alpha in negative feedback is used for initially and defaultly starting all the servers, and then the system starts to synchronously acquire the running state and performance of each server.
Step 102: after the end of each acquisition cycle, adding the CPU utilization of each core to c0And c1In (c), the CPU utilization without packet loss is added to (c)0Will haveCPU utilization of lost packets added to c1. The processing performance is represented by 0 and 1, and whether there is a packet loss or not is 0 if there is no packet loss, and is 1 if there is a packet loss.
Step 103: according to co=argmaxi[Ni 0/Ni 1]0.1 calculate optimal server operational parameters.
Step 103: calculating the average CPU utilization of the current server and recording the average CPU utilization as
Step 104: calculating the number of servers to be started in the next period according to the following formula
Step 105: and intensively scheduling the flow to the M (k +1) servers by using a dynamic load balancing method.
The above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and a person skilled in the art can make modifications or equivalent substitutions to the technical solution of the present invention without departing from the spirit and scope of the present invention, and the scope of the present invention should be determined by the claims.
Claims (4)
1. A server cluster grouping scheduling method based on negative feedback comprises the following steps:
1) calculating the optimal parameters of the server operation in the kth period according to the historical operation state of the server in the kth period; the method for calculating the optimal parameters of the server operation comprises the following steps: setting the CPU utilization rate and the processing performance of the server acquired in the kth period as c ═ c respectively1,c2,…,ci,…,cn],p=[p1,p2,…,pi,…,pn]Wherein n is the total number of servers operating in the k period, ciFor the ith server CPU utilization, piFor the processing performance of the ith server, ci∈(0,1),pi0 or 1; according to piValue taking, namely dividing the set c into two sets c respectively0And c1I.e. the handling properties piAdding server CPU utilization with value 0 to set c0To process the property piAdding server CPU utilization with value 1 to set c1(ii) a Then dividing the CPU utilization rate into a plurality of intervals according to co=argmaxi[Ni 0/Ni 1]H is calculated to obtain the optimal operation parameter c of the k periodo(ii) a Wherein N isi 0Is a set c0The number of samples of the middle CPU utilization rate in the ith interval, Ni 1Is a set c1The number of samples of the middle CPU utilization rate in the ith interval is h, and the h is an interval coefficient;
2) obtaining the number of servers to be started in the k +1 th period in a negative feedback mode according to the optimal parameters of the server operation and the operation state of the server in the k-th period; the method for obtaining the number of servers to be started in the (k +1) th period in a negative feedback mode comprises the following steps: and calculating the difference between the average CPU utilization rate of the server in the k period and the optimal CPU utilization rate in the k period, if the difference is greater than 0, increasing the number of the started servers in the k +1 period, and otherwise, reducing the number of the started servers.
2. The method of claim 1, wherein the processing performance is a packet loss ratio, and when there is a packet loss, the processing performance takes a value of 0, and when there is no packet loss, the processing performance takes a value of 1.
3. The method of claim 1, wherein the interval coefficient h is 0.1.
4. The method of claim 1, wherein a formula is utilizedCalculating the number M (k +1) of servers to be started in the k +1 th period; wherein M (k) is that the k-th period is onThe number of servers,is the average CPU utilization in the k-th cycle, coIs the optimal operating parameter for the k-th cycle and α is the adjustment step in negative feedback.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710416047.XA CN107222540B (en) | 2017-06-06 | 2017-06-06 | Negative feedback-based server cluster grouping scheduling method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710416047.XA CN107222540B (en) | 2017-06-06 | 2017-06-06 | Negative feedback-based server cluster grouping scheduling method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107222540A CN107222540A (en) | 2017-09-29 |
CN107222540B true CN107222540B (en) | 2020-11-20 |
Family
ID=59948295
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710416047.XA Active CN107222540B (en) | 2017-06-06 | 2017-06-06 | Negative feedback-based server cluster grouping scheduling method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107222540B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102833355A (en) * | 2012-09-22 | 2012-12-19 | 广东电子工业研究院有限公司 | Load balance system and cloud computing oriented mechanism |
CN103490956A (en) * | 2013-09-22 | 2014-01-01 | 杭州华为数字技术有限公司 | Self-adaptive energy-saving control method, device and system based on traffic predication |
JP5735899B2 (en) * | 2011-10-25 | 2015-06-17 | 日本電信電話株式会社 | Service providing system, file update method, and distributed management apparatus |
-
2017
- 2017-06-06 CN CN201710416047.XA patent/CN107222540B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5735899B2 (en) * | 2011-10-25 | 2015-06-17 | 日本電信電話株式会社 | Service providing system, file update method, and distributed management apparatus |
CN102833355A (en) * | 2012-09-22 | 2012-12-19 | 广东电子工业研究院有限公司 | Load balance system and cloud computing oriented mechanism |
CN103490956A (en) * | 2013-09-22 | 2014-01-01 | 杭州华为数字技术有限公司 | Self-adaptive energy-saving control method, device and system based on traffic predication |
Also Published As
Publication number | Publication date |
---|---|
CN107222540A (en) | 2017-09-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107196870B (en) | DPDK-based traffic dynamic load balancing method | |
CN107038071B (en) | Storm task flexible scheduling algorithm based on data flow prediction | |
CN113515351B (en) | Resource scheduling implementation method based on energy consumption and QoS (quality of service) cooperative optimization | |
CN108416465B (en) | Workflow optimization method in mobile cloud environment | |
CN110798858A (en) | Distributed task unloading method based on cost efficiency | |
WO2016172848A1 (en) | Method and device for adjusting virtual resources in cloud computing | |
WO2015051685A1 (en) | Task scheduling method, device and system | |
CN110046048A (en) | A kind of load-balancing method adaptively quickly reassigned based on workload | |
CN104360724B (en) | A kind of heat dissipating method of the blade server based on job scheduling | |
CN115718644A (en) | Computing task cross-region migration method and system for cloud data center | |
CN116069503A (en) | Method and system for balancing server load | |
CN110569170A (en) | method, device and equipment for evaluating utilization rate of server and storage medium thereof | |
CN109976873B (en) | Scheduling scheme obtaining method and scheduling method of containerized distributed computing framework | |
CN104796673B (en) | A kind of cloud video monitoring system task cut-in method towards energy optimization | |
Qi et al. | Arena: A Learning-based Synchronization Scheme for Hierarchical Federated Learning--Technical Report | |
CN107222540B (en) | Negative feedback-based server cluster grouping scheduling method | |
Qi et al. | Hwamei: A learning-based synchronization scheme for hierarchical federated learning | |
CN108075915A (en) | A kind of RDMA communication connection pond management methods based on ADAPTIVE CONTROL | |
CN112527498A (en) | Service resource elastic expansion and contraction processing method | |
CN112363827A (en) | Multi-resource index Kubernetes scheduling method based on delay factors | |
CN110850957A (en) | Scheduling method for reducing system power consumption through dormancy in edge computing scene | |
CN108595265B (en) | Intelligent distribution method and system for computing resources | |
CN110865871A (en) | Resource rationalization application-based virtualized cluster resource scheduling method | |
CN107239341A (en) | A kind of resource translation method, system and resources of virtual machine scheduling system | |
CN107277824B (en) | Virtual base station clustering method under C-RAN architecture |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |