CN109960585A - A kind of resource regulating method based on kubernetes - Google Patents

A kind of resource regulating method based on kubernetes Download PDF

Info

Publication number
CN109960585A
CN109960585A CN201910107749.9A CN201910107749A CN109960585A CN 109960585 A CN109960585 A CN 109960585A CN 201910107749 A CN201910107749 A CN 201910107749A CN 109960585 A CN109960585 A CN 109960585A
Authority
CN
China
Prior art keywords
node
pod
priority
score
algorithm
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.)
Granted
Application number
CN201910107749.9A
Other languages
Chinese (zh)
Other versions
CN109960585B (en
Inventor
郑雅羽
韩沈钢
王济浩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN201910107749.9A priority Critical patent/CN109960585B/en
Publication of CN109960585A publication Critical patent/CN109960585A/en
Application granted granted Critical
Publication of CN109960585B publication Critical patent/CN109960585B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5083—Techniques for rebalancing the load in a distributed system
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2209/00—Indexing scheme relating to G06F9/00
    • G06F2209/50—Indexing scheme relating to G06F9/50
    • G06F2209/5021—Priority

Abstract

The present invention relates to one kind to be based onkubernetesResource regulating method, in unalterable rules computing cluster ownNodeScore value generates first node priority query, is obtained with dynamic priority algorithmPodPriority query, two queues filtering cannot dispatchNodeGenerate second node priority query, therefrom select highest priority node withPodPriority query's pop-upPodBinding, binding success enter nextPodSchduling cycle, failure are then preferred from second node priority query using included priority algorithmNodeBinding fails then again without suitableNodeFor thisPodOperation, entrance are nextPodSchduling cycle.The present invention includes static scheduling and dynamic resource load balancing, promotes dispatching efficiency, accelerates task deployment efficiency, improves the load equilibrium of task run globality and entire cluster, actively adjusts the load equilibrium of cluster, improve the level of resources utilization of cluster.

Description

A kind of resource regulating method based on kubernetes
Technical field
The invention belongs to the technical field of the transmission of digital information, such as telegraph communication, in particular to a kind of quickening cluster The resource regulating method based on kubernetes of task deployment efficiency.
Background technique
Kubernetes is a container orchestration engine of Google open source, supports automatically dispose, can stretch on a large scale Contracting, application container management, can manage the state of multiple nodes (Node) and the Pod on node in a large-scale cluster Operation.When disposing an application program in production environment, multiple examples of the application are disposed usually to ask to application Seek carry out load balancing.
In Kubernetes, container (containers) virtualization technology is a kind of sharing mode of server resource, It can satisfy the demand of building customised container on demand, it is more flexible convenient different from traditional virtual technology；In addition, Pod in Kubernetes refers to the set of one or several containers, is the minimum unit of kubernetes deployment, patrols Represent an example of some application on volume, and Kubernetes can manage multiple Pod examples of user's creation, simplify fortune The operation difficulty and operation management cost of dimension personnel.
The load equilibrium of the field of cloud calculation main efficiency for being concerned with scheduling of resource and scheduling of resource.The prior art In, default scheduler selects an optimal node operation Pod example with the pre-selection of node, optimization algorithm, and there is also needles To the preempting priority dispatching algorithm of Pod design, however, the Priorities algorithm in kubernetes default scheduler can root The score value of each Node is calculated according to the requirement of each Pod example, this calculating process reduces the efficiency of scheduling of resource, meanwhile, The sort algorithm of Pod priority query (Pod Priority Queue) in kubernetes uses static priority strategy, This aspect likely results in one big business monopolization part of nodes for a long time, reduces service deployment efficiency, on the other hand very may be used Part low priority Pod can be caused to be unable to run for a long time, influence the operation of whole business.
The patent that patent publication No. is CN107948330A discloses the load based on dynamic priority under a kind of cloud environment Balance policy, Node dynamic priority are improved to dynamic priority algorithm in view of the priority algorithm of Pod in kubernetes The shortcomings that make up static priority in kubernetes；However, the readjustment degree for kubernetes scheduler mainly occurs In the case that, Pod dilatation capacity reducing upgrading abnormal in Pod, Node, Node increase reduction etc., do not send out in cluster stable operation and There is no the loads of dynamic adjustment cluster when raw above-mentioned abnormal to promote cluster load more balanced.
A kind of method of dynamic load leveling is disclosed in the patent that patent publication No. is CN106790726A, but method is only Only all Node are divided to for two high and low load queues, this there are the average load state that part Node has been in cluster, It without being scheduled to Pod, and will appear unnecessary scheduling after this method, reduce the efficiency of system.
Summary of the invention
The present invention solves in the prior art, and kubernetes defaults the Priorities algorithm in scheduler can be according to every The requirement of a Pod example calculates the score value of each Node, reduces the efficiency of scheduling of resource, and the Pod in kubernetes is preferential The sort algorithm of grade queue uses static priority strategy, may cause big business monopolization part of nodes for a long time, reduces business The problem of deployment efficiency, part low priority Pod is unable to run for a long time, influences the operation of whole business, provides a kind of excellent The resource regulating method based on kubernetes changed.
The technical scheme adopted by the invention is that a kind of resource regulating method based on kubernetes, the method packet Include following steps:
Step 1: initializing, the score value of all Node, all Node are added from high to low according to score value in computing cluster First node priority query；Pod all in cluster are monitored；
Step 2: cluster running time T, if the Nodename field of any Pod is sky, by current Pod with dynamic priority Pod priority query, otherwise, return step 1 is added in grade algorithm；
Step 3: the pre-selection that the high priority Pod of priority match and first node priority query are passed through into kubernetes Algorithm filtering useless Node；
Step 4: if dispatching and failing without the Node for meeting Pod operation demand, return step 3 enters the tune of next Pod Degree circulation；If it exists can Node, then generate and filter later second node priority query, carry out in next step；
Step 5: the Node of highest priority and the Gao You of the priority match are selected from second node priority query First grade Pod carries out bindings；
Step 6: if binding success, the high priority Pod is operated on the Node selected, under return step 3 enters The schduling cycle of one Pod, otherwise, Bind Failed carries out in next step；
Step 7: the preferred Node of priority algorithm of kubernetes is utilized from second node priority query；With preferred The high priority Pod of Node and the priority match carries out bindings；
Step 8: if binding success, the high priority Pod is operated on the Node selected, under return step 3 enters The schduling cycle of one Pod, otherwise, Bind Failed still enters the schduling cycle of next Pod, until according in cluster Close the Node that Pod to be dispatched is required.
Preferably, in the step 1, calculate the score value of all nodes the following steps are included:
Step 1.1: the score value score of all nodes is calculated with minimal consumption algorithm1, score1Including cpu utilization rate, interior Deposit the sum of utilization rate and network bandwidth utilization factor；
Step 1.2: the score value score of all nodes is calculated with resource most equalization algorithm2；
Step 1.3: with score value score1And score2It is added, obtains the total score score of any node.
Preferably, in the step 1.1, score1=cpu ((capacity-sum (requested)) 10/ capacity)+memory((capacity-sum(requested))10/capacity)+network((cap acity-sum (requested)) 10/capacity), wherein first item is cpu utilization rate, Section 2 is memory usage, Section 3 is net Network bandwidth availability ratio, capacity indicate total amount of the every kind of resource on each node, and sum (requested) indicates the Pod The corresponding total amount of required resource.
Preferably, in the step 2, dynamic priority algorithm the following steps are included:
Step 2.1: setting initial priority value initPodPriorityValue, system high priority threshold alpha, system are low Priority threshold value β, Pod flees from listening period Tescape, the desired Pod quantity of minimum needed for each Pod service operation MinPodAmount, Pod run quantity weight W1, occupy the time weight W of Pod priority query2, Pod runing time weight W3；
Step 2.2: being greater than T when the time that Pod is established in waitingescape, carry out in next step；
Step 2.3: operation quantity runningPodAmount, corresponding the Pod expectation for obtaining each Pod run quantity NeededPodAmount, the time T for occupying Pod priority queryqueue, the position Num in queue and operation time Truntime；
Step 2.4: the priority value runtimePodPriorityValue of each Pod in Pod priority query is calculated, RuntimePodPriorityValue=W1*runningPodAmount+W2*Num*Tqueue+W3*Truntime；
Step 2.5: if runtimePodPriorityValue is greater than α, being entered from large to small according to value reduces priority Queue, calculate decreasePodPriorityValue, decreasePodPriorityValue= initPodPriorityValue*(1-runningPodAmount/neededPodAmount)；It carries out in next step；
If runtimePodPriorityValue is less than β, enter the queue for increasing priority from small to large according to value, Calculate increasePodPriorityValue, increasePodPriorityValue=initPodPriorityValue* (1+minPodAmount/nee dedPodAmount)；It carries out in next step；
If runtimePodPriorityValue is between threshold alpha and β, without the adjustment of dynamic priority, carry out Step 3；
Step 2.6: all Pod being rejoined into Pod priority query, update the information of cluster.
Preferably, in the step 2.5, if decreasePodPriorityValue is not less than α, corresponding Pod is reduced Priority；If increasePodPriorityValue is not more than β, increase the priority of corresponding Pod.
Preferably, in the step 5, bindings are to be changed to the value of the Nodename field of the Pod selected to select The name of Node.
Preferably, in the step 7, priority algorithm include LeastRequestedPriority algorithm, BalanceResourceAllocation algorithm, SelectorSpreadPriority algorithm, NodeAffinityPriority Algorithm, TaintTolerationPriority algorithm and InterPodAffinityPriority algorithm, any Node, which meets, to be appointed One algorithm is then multiplied with the algorithm with corresponding weight coefficient, and all product additions obtain the score value of Node, with score value Higher Node is that priority is higher.
Preferably, the resource regulating method further includes cluster dynamic resource load-balancing method, and the cluster dynamic provides Source load-balancing method the following steps are included:
Step 9.1: the threshold value η of initialization high load Node1With the threshold value η of low-load Node2；
Step 9.2: monitoring cluster information in real time, according to certain polling cycle T, obtain the status information of all Node；
Step 9.3: calculating score value Avg_Score (i)=(1-avg_cpu (i)) * (1- of all Node average loads Avg_network (i)) * (1-avg_storage (i)), calculate load score value Score (i)=(1-cpu of each Node (i)) * (1-network (i)) * (1-storage (i)), wherein i is Node serial number, avg_cpu (i), avg_network (i) With avg_storage (i) be respectively the CPU of all Node in cluster, network bandwidth, memory usage average value, cpu (i), Network (i) and storage (i) is respectively the CPU, network bandwidth, memory usage of each Node；
Step 9.4: enabling η1=λ1Avg_Score(i)、η2=λ2Avg_Score(i)；
Step 9.5: according to the Score (i) of each node, with η1And η2As threshold value, by Score (i) < η1Node return For high load queue, by Score (i) > η2Node be classified as low-load queue, by η1≤Score(i)≤η2Node be classified as Weigh load queue；
Step 9.6: if high load queue and low-load queue are not sky, carrying out in next step；Otherwise without dynamic Load dispatch, return step 9.2 enter next polling cycle；
Step 9.7: the Pod run on the smallest Node of score value in selection high load queue is preselected in low-load queue Pod is run to the Node in low-load queue, reaches cluster load balance by Node.
Preferably, in the step 9.2, the status information of Node includes the cpu busy percentage of Node, memory usage, net Network bandwidth availability ratio and space utilisation.
Preferably, in the step 9.3,
The present invention provides a kind of resource regulating methods based on kubernetes of optimization, by reading cluster information, The score value of all Node is calculated with unalterable rules, generates first node priority query, while obtaining and calculating according to dynamic priority The Pod priority query that method obtains, two queues filter out the Node that cannot be dispatched by preselecting algorithm, and it is excellent to generate second node First grade queue selects the node of highest priority and Pod priority query to pop up directly from second node priority query Pod binding, binding success then enter the schduling cycle of next Pod, and it is preferential that Bind Failed then uses kubernetes to carry Grade algorithm preferred Node from second node priority query is bound, and binding success is then followed into the scheduling of next Pod Ring, scheduling failure, which then represents cluster, does not have suitable Node that can run for the Pod, into the schduling cycle of next Pod.
The present invention includes static scheduling and dynamic resource load balancing, and selection one is appropriate and is not that optimal node can To save the scheduling time of kubernetes scheduler, the efficiency of its scheduling is improved, the task deployment efficiency of cluster is accelerated, The globality of task run is improved, and improves the load equilibrium of entire cluster；Dynamic load balancing method is negative by height Pod on load Node, which is transferred on low-load Node, to be run, and is actively adjusted the load equilibrium of cluster, is improved the resource of cluster Utilization efficiency.
Detailed description of the invention
Fig. 1 is flow chart of the invention；
Fig. 2 is the flow chart of dynamic priority algorithm in the present invention；
Fig. 3 is the flow chart of cluster dynamic resource load-balancing method in the present invention.
Specific embodiment
The present invention is described in further detail below with reference to case study on implementation, but practical range of the invention is not limited to This.
The present invention relates to a kind of resource regulating methods based on kubernetes, including static scheduling and dynamic resource load It is balanced.
It the described method comprises the following steps.
Step 1: initializing, the score value of all Node, all Node are added from high to low according to score value in computing cluster First node priority query；Pod all in cluster are monitored.
In the step 1, calculate the score value of all nodes the following steps are included:
Step 1.1: the score value score of all nodes is calculated with minimal consumption algorithm1, score1Including cpu utilization rate, interior Deposit the sum of utilization rate and network bandwidth utilization factor；
In the step 1.1, score1=cpu ((capacity-sum (requested)) 10/capacity)+ memory((capacity-sum(requested))10/capacity)+network((cap acity-sum (requested)) 10/capacity), wherein first item is cpu utilization rate, Section 2 is memory usage, Section 3 is net Network bandwidth availability ratio, capacity indicate total amount of the every kind of resource on each node, and sum (requested) indicates the Pod The corresponding total amount of required resource.
Step 1.2: the score value score of all nodes is calculated with resource most equalization algorithm2；
Step 1.3: with score value score1And score2It is added, obtains the total score score of any node.
In the present invention, two algorithms that initialization calculates the score value of Node node include minimal consumption algorithm LeastRequestedPriority and resource most equalization algorithm BalancedResourceAllocation, the former is for selecting The smallest node of resource consumption, resource herein include cpu busy percentage, memory usage, network bandwidth utilization factor, the latter's selection Resource uses most balanced node, is algorithm known, the mainly equilibrium between memory and cpu busy percentage.
In the present invention, score2=10-variance (cpuFraction, memoryFraction) * 10, wherein CpuFraction and memoryFraction is the ratio of available resources on Pod request resource and Node, cpuFraction= Cpu (requested)/cpu (available), memoryFraction=memory (requested)/memory (available), variance is balanced algorithm between the included calculating two kinds of resources of kubernetes, requested Indicate demand, available indicates Availability.
Step 2: cluster running time T, if the Nodename field of any Pod is sky, by current Pod with dynamic priority Pod priority query, otherwise, return step 1 is added in grade algorithm.
In the present invention, the dynamic priority algorithm master of Pod is to solve high priority Pod, and to occupy Pod for a long time preferential The Head-of-line of grade queue, causes high priority Pod that may occupy the Node resource of cluster for a long time, and low priority occurs The case where Pod task is unable to run.Specifically, dynamic priority algorithm refers to the calculation run after cluster operation a period of time Method can prevent the operation of cluster one from occurring as soon as the case where height priority P od priority changes, and high priority Pod is excellent First grade reduces, and low priority Pod priority increases.
In the present invention, when problematic there is no Pod, then back to the state for monitoring Pod, continue point for calculating Node Value.
In the step 2, dynamic priority algorithm the following steps are included:
Step 2.1: setting initial priority value initPodPriorityValue, system high priority threshold alpha, system are low Priority threshold value β, Pod flees from listening period Tescape, the desired Pod quantity of minimum needed for each Pod service operation MinPodAmount, Pod run quantity weight W1, occupy the time weight W of Pod priority query2, Pod runing time weight W3；
Step 2.2: being greater than T when the time that Pod is established in waitingescape, carry out in next step；
Step 2.3: operation quantity runningPodAmount, corresponding the Pod expectation for obtaining each Pod run quantity NeededPodAmount, the time T for occupying Pod priority queryqueue, the position Num in queue and operation time Truntime；
Step 2.4: the priority value runtimePodPriorityValue of each Pod in Pod priority query is calculated, RuntimePodPriorityValue=W1*runningPodAmount+W2*Num*Tqueue+W3*Truntime；
Step 2.5: if runtimePodPriorityValue is greater than α, being entered from large to small according to value reduces priority Queue, calculate decreasePodPriorityValue, decreasePodPriorityValue= initPodPriorityValue*(1-runningPodAmount/neededPodAmount)；It carries out in next step；
If runtimePodPriorityValue is less than β, enter the queue for increasing priority from small to large according to value, Calculate increasePodPriorityValue, increasePodPriorityValue=initPodPriorityValue* (1+minPodAmount/nee dedPodAmount)；It carries out in next step；
If runtimePodPriorityValue is between threshold alpha and β, without the adjustment of dynamic priority, carry out Step 3；
In the step 2.5, if decreasePodPriorityValue is not less than α, the preferential of corresponding Pod is reduced Grade；If increasePodPriorityValue is not more than β, increase the priority of corresponding Pod.
The time for establishing Pod is waited to be greater than T in the present invention, in step 2.2escapeIt is to indicate that Pod is created successfully.
In the present invention, decreasePodPriorityValue and increasePodPriorityValue are further judged So that not having to the whole sequence for the priority for changing all Pod, the priority of all Pod is prevented to be all located between α and β, thus Scheduling of resource is caused the case where congestion occur.
Step 2.6: all Pod being rejoined into Pod priority query, update the information of cluster.
Step 3: the pre-selection that the high priority Pod of priority match and first node priority query are passed through into kubernetes Algorithm filtering useless Node.
In the present invention, so-called useless Node is obvious undesirable node, is including but not limited to filtered out The upper residue CPU of Node and memory source are inadequate, persistent storage volume presence conflict, there are stain to cause by host Node It can not dispatch, Pod affine and anti-affine etc. be unsatisfactory for desired Node.
Step 4: if dispatching and failing without the Node for meeting Pod operation demand, return step 3 enters the tune of next Pod Degree circulation；If it exists can Node, then generate and filter later second node priority query, carry out in next step.
In the present invention, meets Pod operation demand and refer to that memory, bandwidth, the CPU quantity etc. that meet Pod operation require.
Step 5: the Node of highest priority and the Gao You of the priority match are selected from second node priority query First grade Pod carries out bindings.
In the step 5, bindings are that the value of the Nodename field of the Pod selected is changed to the name of the Node selected Word.
In the present invention, Kubernetes only will be updated Pod the and Node information in caching, finally selecting in the binding stage It can verify whether the Pod is determined to operate on the Node again on Node out.
In the present invention, Pod is popped up, the Nodename of the Node selected is written to the attribute value of the Nodename in Pod, is made For binding.
Step 6: if binding success, the high priority Pod is operated on the Node selected, under return step 3 enters The schduling cycle of one Pod, otherwise, Bind Failed carries out in next step.
Step 7: the preferred Node of priority algorithm of kubernetes is utilized from second node priority query；With preferred The high priority Pod of Node and the priority match carries out bindings.
In the step 7, priority algorithm include LeastRequestedPriority algorithm, BalanceResourceAllocation algorithm, SelectorSpreadPriority algorithm, NodeAffinityPriority Algorithm, TaintTolerationPriority algorithm and InterPodAffinityPriority algorithm, any Node, which meets, to be appointed One algorithm is then multiplied with the algorithm with corresponding weight coefficient, and all product additions obtain the score value of Node, with score value Higher Node is that priority is higher.
In the present invention, the score value of all Node in second node priority query, a Node are calculated by these algorithms The field number for meeting above-mentioned rule is more, then score is higher.
Step 8: if binding success, the high priority Pod is operated on the Node selected, under return step 3 enters The schduling cycle of one Pod, otherwise, Bind Failed still enters the schduling cycle of next Pod, until according in cluster Close the Node that Pod to be dispatched is required.
In the present invention, bindings are identical twice.
The resource regulating method further includes cluster dynamic resource load-balancing method, and the cluster dynamic resource load is equal Weighing apparatus method includes the following steps.
Step 9.1: the threshold value η of initialization high load Node1With the threshold value η of low-load Node2。
In the present invention, η1And η2According to the load mean value of Servers-all by those skilled in the art's self-setting.
Step 9.2: monitoring cluster information in real time, according to certain polling cycle T, obtain the status information of all Node.
In the step 9.2, the status information of Node includes the cpu busy percentage of Node, memory usage, network bandwidth benefit With rate and space utilisation.
Step 9.3: calculating score value Avg_Score (i)=(1-avg_cpu (i)) * (1- of all Node average loads Avg_network (i)) * (1-avg_storage (i)), calculate load score value Score (i)=(1-cpu of each Node (i)) * (1-network (i)) * (1-storage (i)), wherein i is Node serial number, avg_cpu (i), avg_network (i) With avg_storage (i) be respectively the CPU of all Node in cluster, network bandwidth, memory usage average value, cpu (i), Network (i) and storage (i) is respectively the CPU, network bandwidth, memory usage of each Node.
In the step 9.3,
Step 9.4: enabling η1=λ1Avg_Score(i)、η2=λ2Avg_Score(i)。
Step 9.5: according to the Score (i) of each node, with η1And η2As threshold value, by Score (i) < η1Node return For high load queue, by Score (i) > η2Node be classified as low-load queue, by η1≤Score(i)≤η2Node be classified as Weigh load queue.
Step 9.6: if high load queue and low-load queue are not sky, carrying out in next step；Otherwise without dynamic Load dispatch, return step 9.2 enter next polling cycle.
Step 9.7: the Pod run on the smallest Node of score value in selection high load queue is preselected in low-load queue Pod is run to the Node in low-load queue, reaches cluster load balance by Node.
In the present invention, 0 < λ1< λ2。
In the present invention, η1< η2。
The Node preselected in low-load queue in the present invention, in step 9.7 is included known using kubernetes Algorithm.
The present invention is calculated the score value of all Node with unalterable rules, it is preferential to generate first node by reading cluster information Grade queue, while the Pod priority query obtained according to dynamic priority algorithm is obtained, two queues pass through pre-selection algorithm filtering Fall the Node that cannot be dispatched, generates second node priority query, directly select priority from second node priority query Highest node and the Pod of Pod priority query pop-up are bound, and binding success then enters the schduling cycle of next Pod, binds The priority algorithm that failure is then carried using kubernetes preferred Node from second node priority query is bound, and is tied up Fixed success then enters the schduling cycle of next Pod, and scheduling failure, which then represents cluster, does not have suitable Node that can transport for the Pod Row, into the schduling cycle of next Pod.
The present invention includes static scheduling and dynamic resource load balancing, and selection one is appropriate and is not that optimal node can To save the scheduling time of kubernetes scheduler, the efficiency of its scheduling is improved, the task deployment efficiency of cluster is accelerated, The globality of task run is improved, and improves the load equilibrium of entire cluster；Dynamic load balancing method is negative by height Pod on load Node, which is transferred on low-load Node, to be run, and is actively adjusted the load equilibrium of cluster, is improved the resource of cluster Utilization efficiency.

Claims (10)

1. a kind of resource regulating method based on kubernetes, it is characterised in that: the described method comprises the following steps:
Step 1: initializing, the score value of all Node in computing cluster, all Node are added first from high to low according to score value Node priority queue；Pod all in cluster are monitored；
Step 2: cluster running time T is calculated current Pod with dynamic priority if the Nodename field of any Pod is sky Pod priority query, otherwise, return step 1 is added in method；
Step 3: the high priority Pod of priority match and first node priority query are passed through to the pre-selection algorithm of kubernetes Filtering useless Node；
Step 4: if dispatching and failing without the Node for meeting Pod operation demand, the scheduling that return step 3 enters next Pod is followed Ring；If it exists can Node, then generate and filter later second node priority query, carry out in next step；
Step 5: the Node of highest priority and the high priority of the priority match are selected from second node priority query Pod carries out bindings；
Step 6: if binding success, the high priority Pod is operated on the Node selected, and return step 3 enters next The schduling cycle of Pod, otherwise, Bind Failed carries out in next step；
Step 7: the preferred Node of priority algorithm of kubernetes is utilized from second node priority query；With preferred Node Bindings are carried out with the high priority Pod of the priority match；
Step 8: if binding success, the high priority Pod is operated on the Node selected, and return step 3 enters next The schduling cycle of Pod, otherwise, Bind Failed still enter the schduling cycle of next Pod, until occur meeting in cluster to Dispatch the Node that Pod is required.
2. a kind of resource regulating method based on kubernetes according to claim 1, it is characterised in that: the step In 1, calculate the score value of all nodes the following steps are included:
Step 1.1: the score value score of all nodes is calculated with minimal consumption algorithm1, score1Including cpu utilization rate, memory benefit With the sum of rate and network bandwidth utilization factor；
Step 1.2: the score value score of all nodes is calculated with resource most equalization algorithm2；
Step 1.3: with score value score1And score2It is added, obtains the total score score of any node.
3. a kind of resource regulating method based on kubernetes according to claim 2, it is characterised in that: the step In 1.1, score1=cpu ((capacity-sum (requested)) 10/capacity)+memory ((capacity-sum (requested)) 10/capacity)+network ((capacity-sum (requested)) 10/capacity), wherein First item is cpu utilization rate, Section 2 is memory usage, Section 3 is network bandwidth utilization factor, and capacity indicates every kind Total amount of the resource on each node, sum (requested) indicate the corresponding total amount of the required resource of Pod.
4. a kind of resource regulating method based on kubernetes according to claim 1, it is characterised in that: the step In 2, dynamic priority algorithm the following steps are included:
Step 2.1: setting initial priority value initPodPriorityValue, system high priority threshold alpha, system are low preferential Grade threshold value beta, Pod flee from listening period Tescape, the desired Pod quantity of minimum needed for each Pod service operation MinPodAmount, Pod run quantity weight W1, occupy the time weight W of Pod priority query2, Pod runing time weight W3；
Step 2.2: being greater than T when the time that Pod is established in waitingescape, carry out in next step；
Step 2.3: operation quantity runningPodAmount, corresponding the Pod expectation for obtaining each Pod run quantity NeededPodAmount, the time T for occupying Pod priority queryqueue, the position Num in queue and operation time Truntime；
Step 2.4: the priority value runtimePodPriorityValue of each Pod in Pod priority query is calculated, RuntimePodPriorityValue=W1*runningPodAmount+W2*Num*Tqueue+W3*Truntime；
Step 2.5: if runtimePodPriorityValue is greater than α, entering the team for reducing priority from large to small according to value Column calculate decreasePodPriorityValue, decreasePodPriorityValue= initPodPriorityValue*(1-runningPodAmount/neededPodAmount)；It carries out in next step；
If runtimePodPriorityValue is less than β, enter the queue for increasing priority from small to large according to value, calculates IncreasePodPriorityValue, increasePodPriorityValue=initPodPriorityValue* (1+ minPodAmount/neededPodAmount)；It carries out in next step；
If runtimePodPriorityValue is between threshold alpha and β, without the adjustment of dynamic priority, step is carried out 3；
Step 2.6: all Pod being rejoined into Pod priority query, update the information of cluster.
5. a kind of resource regulating method based on kubernetes according to claim 4, it is characterised in that: the step In 2.5, if decreasePodPriorityValue is not less than α, the priority of corresponding Pod is reduced；If IncreasePodPriorityValue is not more than β, then increases the priority of corresponding Pod.
6. a kind of resource regulating method based on kubernetes according to claim 1, it is characterised in that: the step In 5, bindings are that the value of the Nodename field of the Pod selected is changed to the name of the Node selected.
7. a kind of resource regulating method based on kubernetes stated according to claim 1, it is characterised in that: the step 7 In, priority algorithm include LeastRequestedPriority algorithm, BalanceResourceAllocation algorithm, SelectorSpreadPriority algorithm, NodeAffinityPriority algorithm, TaintTolerationPriority are calculated Method and InterPodAffinityPriority algorithm, any Node meet any algorithm, then with the algorithm and corresponding power Value coefficient is multiplied, and all product additions obtain the score value of Node, higher as priority using the higher Node of score value.
8. a kind of resource regulating method based on kubernetes according to claim 1, it is characterised in that: the resource Dispatching method further includes cluster dynamic resource load-balancing method, and the cluster dynamic resource load-balancing method includes following step It is rapid:
Step 9.1: the threshold value η of initialization high load Node1With the threshold value η of low-load Node2；
Step 9.2: monitoring cluster information in real time, according to certain polling cycle T, obtain the status information of all Node；
Step 9.3: calculating score value Avg_Score (i)=(1-avg_cpu (i)) * (1-avg_ of all Node average loads Network (i)) * (1-avg_storage (i)), calculate load score value Score (i)=(1-cpu (i)) * (1- of each Node Network (i)) * (1-storage (i)), wherein i is Node serial number, avg_cpu (i), avg_network (i) and avg_ Storage (i) be respectively the CPU of all Node in cluster, network bandwidth, memory usage average value, cpu (i), Network (i) and storage (i) is respectively the CPU, network bandwidth, memory usage of each Node；
Step 9.4: enabling η1=λ1Avg_Score(i)、η2=λ2Avg_Score(i)；
Step 9.5: according to the Score (i) of each node, with η1And η2As threshold value, by Score (i) < η1Node be classified as height Load queue, by Score (i) > η2Node be classified as low-load queue, by η1≤Score(i)≤η2Node be classified as it is balanced negative Carry queue；
Step 9.6: if high load queue and low-load queue are not sky, carrying out in next step；Otherwise without dynamic load Scheduling, return step 9.2 enter next polling cycle；
Step 9.7: the Pod run on the smallest Node of score value in selection high load queue preselects the Node in low-load queue, Pod is run into the Node in low-load queue, reaches cluster load balance.
9. a kind of resource regulating method based on kubernetes according to claim 8, it is characterised in that: the step In 9.2, the status information of Node includes that the cpu busy percentage of Node, memory usage, network bandwidth utilization factor and storage utilize Rate.
10. a kind of resource regulating method based on kubernetes according to claim 8, it is characterised in that: the step In rapid 9.3,
CN201910107749.9A 2019-02-02 2019-02-02 Resource scheduling method based on kubernets Active CN109960585B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910107749.9A CN109960585B (en) 2019-02-02 2019-02-02 Resource scheduling method based on kubernets

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910107749.9A CN109960585B (en) 2019-02-02 2019-02-02 Resource scheduling method based on kubernets

Publications (2)

Publication Number Publication Date
CN109960585A true CN109960585A (en) 2019-07-02
CN109960585B CN109960585B (en) 2021-05-14

Family

ID=67023606

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910107749.9A Active CN109960585B (en) 2019-02-02 2019-02-02 Resource scheduling method based on kubernets

Country Status (1)

Country Link
CN (1) CN109960585B (en)

Cited By (39)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110515730A (en) * 2019-08-22 2019-11-29 北京宝兰德软件股份有限公司 Resource secondary dispatching method and device based on kubernetes container arranging system
CN110704165A (en) * 2019-10-08 2020-01-17 浪潮云信息技术有限公司 High-availability deployment method for container
CN110719335A (en) * 2019-10-21 2020-01-21 中国科学院空间应用工程与技术中心 Resource scheduling method, system and storage medium under space-based cloud computing architecture
CN110727653A (en) * 2019-10-12 2020-01-24 广州华多网络科技有限公司 Multi-project load balancing method and device
CN110851236A (en) * 2019-11-11 2020-02-28 星环信息科技（上海）有限公司 Real-time resource scheduling method and device, computer equipment and storage medium
CN110941393A (en) * 2019-10-31 2020-03-31 北京浪潮数据技术有限公司 Logical volume management-based LV supply method, device, equipment and medium
CN110990121A (en) * 2019-11-28 2020-04-10 中国—东盟信息港股份有限公司 Kubernetes scheduling strategy based on application portrait
CN111090523A (en) * 2019-12-18 2020-05-01 浪潮云信息技术有限公司 Resource scheduling method based on dynamic priority under kubernets environment
CN111143059A (en) * 2019-12-17 2020-05-12 天津大学 Improved Kubernetes resource scheduling method
CN111258609A (en) * 2020-01-19 2020-06-09 北京百度网讯科技有限公司 Upgrading method and device of Kubernetes cluster, electronic equipment and medium
CN111352717A (en) * 2020-03-24 2020-06-30 广西梯度科技有限公司 Method for realizing kubernets self-defined scheduler
CN111475251A (en) * 2020-03-08 2020-07-31 苏州浪潮智能科技有限公司 Cluster container scheduling method, system, terminal and storage medium
CN111522639A (en) * 2020-04-16 2020-08-11 南京邮电大学 Multidimensional resource scheduling method under Kubernetes cluster architecture system
CN111694646A (en) * 2020-05-29 2020-09-22 北京百度网讯科技有限公司 Resource scheduling method and device, electronic equipment and computer readable storage medium
CN111737003A (en) * 2020-06-24 2020-10-02 重庆紫光华山智安科技有限公司 Pod balanced scheduling method and device, main node and storage medium
CN111813556A (en) * 2020-07-21 2020-10-23 北京东方通软件有限公司 Elastic expansion method of virtual cluster in cloud computing environment
CN112181661A (en) * 2020-10-13 2021-01-05 极道科技（北京）有限公司 Task scheduling method
CN112214303A (en) * 2019-07-09 2021-01-12 上海交通大学 Kubernetes cluster automatic scaling system
CN112214288A (en) * 2019-07-10 2021-01-12 中国移动通信集团上海有限公司 Pod scheduling method, device, equipment and medium based on Kubernetes cluster
CN112286631A (en) * 2020-10-23 2021-01-29 烽火通信科技股份有限公司 Kubernetes resource scheduling method and device and electronic equipment
CN112286671A (en) * 2020-12-29 2021-01-29 湖南星河云程信息科技有限公司 Containerization batch processing job scheduling method and device and computer equipment
CN112363811A (en) * 2020-11-16 2021-02-12 中国电子科技集团公司电子科学研究院 Artificial intelligence computing resource scheduling method and computer readable storage medium
CN112527454A (en) * 2020-12-04 2021-03-19 上海连尚网络科技有限公司 Container group scheduling method and device, electronic equipment and computer readable medium
WO2021063339A1 (en) * 2019-09-30 2021-04-08 星环信息科技(上海)股份有限公司 Cluster resource scheduling method, apparatus, device and storage medium
CN112764906A (en) * 2021-01-26 2021-05-07 浙江工业大学 Cluster resource scheduling method based on user job type and node performance bias
CN112988380A (en) * 2021-02-25 2021-06-18 电子科技大学 Kubernetes-based cluster load adjusting method and storage medium
WO2021120633A1 (en) * 2019-12-19 2021-06-24 华为技术有限公司 Load balancing method and related device
CN113391891A (en) * 2021-05-20 2021-09-14 国网江苏省电力有限公司信息通信分公司 Load balancing resource scheduling method based on Rete and character string pattern matching algorithm
CN113485792A (en) * 2021-07-08 2021-10-08 厦门服云信息科技有限公司 Pod scheduling method in kubernets cluster, terminal equipment and storage medium
CN114064296A (en) * 2022-01-18 2022-02-18 北京建筑大学 Kubernetes scheduling method, Kubernetes scheduling device and storage medium
CN114327851A (en) * 2020-09-30 2022-04-12 慧与发展有限责任合伙企业 Preventing scheduling or execution of resources on inconsistent host nodes
CN114327886A (en) * 2021-12-24 2022-04-12 国家石油天然气管网集团有限公司 Dynamic resource scheduling method based on big data deep learning
US11336588B2 (en) 2020-06-26 2022-05-17 Red Hat, Inc. Metadata driven static determination of controller availability
CN114625493A (en) * 2020-12-14 2022-06-14 中国石油大学(华东) Kubernetes cluster resource scheduling method based on improved intelligent longicorn stigma method
US11474905B2 (en) 2020-12-10 2022-10-18 International Business Machines Corporation Identifying harmful containers
CN115297112A (en) * 2022-07-31 2022-11-04 南京匡吉信息科技有限公司 Dynamic resource quota and scheduling component based on Kubernetes
US11706105B2 (en) 2021-12-14 2023-07-18 International Business Machines Corporation Selecting low priority pods for guaranteed runs
CN117112239A (en) * 2023-10-23 2023-11-24 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Extensible load balancing method and system on heterogeneous reasoning back end
WO2024021467A1 (en) * 2022-07-26 2024-02-01 天翼云科技有限公司 Cluster resource planning method, device, apparatus, and medium

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106027643A (en) * 2016-05-18 2016-10-12 无锡华云数据技术服务有限公司 Resource scheduling method based on Kubernetes container cluster management system
CN106886451A (en) * 2017-01-10 2017-06-23 广东石油化工学院 A kind of Multi-workflow method for allocating tasks based on virtualization container technique
CN108769254A (en) * 2018-06-25 2018-11-06 星环信息科技（上海）有限公司 Resource-sharing application method, system and equipment based on preemption scheduling
CN108829509A (en) * 2018-05-03 2018-11-16 山东汇贸电子口岸有限公司 Distributed container cluster framework resources management method based on domestic CPU and operating system
US20180351851A1 (en) * 2016-12-15 2018-12-06 Nutanix, Inc. Managing validity periods for computing resource attributes
CN109117265A (en) * 2018-07-12 2019-01-01 北京百度网讯科技有限公司 The method, apparatus, equipment and storage medium of schedule job in the cluster
CN109167835A (en) * 2018-09-13 2019-01-08 重庆邮电大学 A kind of physics resource scheduling method and system based on kubernetes
US10191778B1 (en) * 2015-11-16 2019-01-29 Turbonomic, Inc. Systems, apparatus and methods for management of software containers
CN109298914A (en) * 2018-10-12 2019-02-01 西安交通大学 A kind of Docker based on three-tier architecture and virtual machine initial placement method

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10191778B1 (en) * 2015-11-16 2019-01-29 Turbonomic, Inc. Systems, apparatus and methods for management of software containers
CN106027643A (en) * 2016-05-18 2016-10-12 无锡华云数据技术服务有限公司 Resource scheduling method based on Kubernetes container cluster management system
US20180351851A1 (en) * 2016-12-15 2018-12-06 Nutanix, Inc. Managing validity periods for computing resource attributes
CN106886451A (en) * 2017-01-10 2017-06-23 广东石油化工学院 A kind of Multi-workflow method for allocating tasks based on virtualization container technique
CN108829509A (en) * 2018-05-03 2018-11-16 山东汇贸电子口岸有限公司 Distributed container cluster framework resources management method based on domestic CPU and operating system
CN108769254A (en) * 2018-06-25 2018-11-06 星环信息科技（上海）有限公司 Resource-sharing application method, system and equipment based on preemption scheduling
CN109117265A (en) * 2018-07-12 2019-01-01 北京百度网讯科技有限公司 The method, apparatus, equipment and storage medium of schedule job in the cluster
CN109167835A (en) * 2018-09-13 2019-01-08 重庆邮电大学 A kind of physics resource scheduling method and system based on kubernetes
CN109298914A (en) * 2018-10-12 2019-02-01 西安交通大学 A kind of Docker based on three-tier architecture and virtual machine initial placement method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
唐瑞: "基于Kubernetes的容器云平台资源调度策略研究", 《中国优秀硕士学位论文数据库》 *

Cited By (54)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112214303A (en) * 2019-07-09 2021-01-12 上海交通大学 Kubernetes cluster automatic scaling system
CN112214288A (en) * 2019-07-10 2021-01-12 中国移动通信集团上海有限公司 Pod scheduling method, device, equipment and medium based on Kubernetes cluster
CN112214288B (en) * 2019-07-10 2023-04-25 中国移动通信集团上海有限公司 Pod scheduling method, device, equipment and medium based on Kubernetes cluster
CN110515730A (en) * 2019-08-22 2019-11-29 北京宝兰德软件股份有限公司 Resource secondary dispatching method and device based on kubernetes container arranging system
WO2021063339A1 (en) * 2019-09-30 2021-04-08 星环信息科技(上海)股份有限公司 Cluster resource scheduling method, apparatus, device and storage medium
CN110704165A (en) * 2019-10-08 2020-01-17 浪潮云信息技术有限公司 High-availability deployment method for container
CN110727653A (en) * 2019-10-12 2020-01-24 广州华多网络科技有限公司 Multi-project load balancing method and device
CN110727653B (en) * 2019-10-12 2023-06-16 广州华多网络科技有限公司 Multi-project load balancing method and device
CN110719335A (en) * 2019-10-21 2020-01-21 中国科学院空间应用工程与技术中心 Resource scheduling method, system and storage medium under space-based cloud computing architecture
CN110941393A (en) * 2019-10-31 2020-03-31 北京浪潮数据技术有限公司 Logical volume management-based LV supply method, device, equipment and medium
WO2021093783A1 (en) * 2019-11-11 2021-05-20 星环信息科技(上海)股份有限公司 Real-time resource scheduling method and apparatus, computer device, and storage medium
CN110851236A (en) * 2019-11-11 2020-02-28 星环信息科技（上海）有限公司 Real-time resource scheduling method and device, computer equipment and storage medium
CN110990121A (en) * 2019-11-28 2020-04-10 中国—东盟信息港股份有限公司 Kubernetes scheduling strategy based on application portrait
CN111143059B (en) * 2019-12-17 2023-10-20 天津大学 Improved Kubernetes resource scheduling method
CN111143059A (en) * 2019-12-17 2020-05-12 天津大学 Improved Kubernetes resource scheduling method
CN111090523A (en) * 2019-12-18 2020-05-01 浪潮云信息技术有限公司 Resource scheduling method based on dynamic priority under kubernets environment
WO2021120633A1 (en) * 2019-12-19 2021-06-24 华为技术有限公司 Load balancing method and related device
CN111258609A (en) * 2020-01-19 2020-06-09 北京百度网讯科技有限公司 Upgrading method and device of Kubernetes cluster, electronic equipment and medium
CN111475251A (en) * 2020-03-08 2020-07-31 苏州浪潮智能科技有限公司 Cluster container scheduling method, system, terminal and storage medium
CN111352717A (en) * 2020-03-24 2020-06-30 广西梯度科技有限公司 Method for realizing kubernets self-defined scheduler
CN111352717B (en) * 2020-03-24 2023-04-07 广西梯度科技股份有限公司 Method for realizing kubernets self-defined scheduler
CN111522639B (en) * 2020-04-16 2022-11-01 南京邮电大学 Multidimensional resource scheduling method under Kubernetes cluster architecture system
CN111522639A (en) * 2020-04-16 2020-08-11 南京邮电大学 Multidimensional resource scheduling method under Kubernetes cluster architecture system
CN111694646A (en) * 2020-05-29 2020-09-22 北京百度网讯科技有限公司 Resource scheduling method and device, electronic equipment and computer readable storage medium
CN111694646B (en) * 2020-05-29 2023-11-07 北京百度网讯科技有限公司 Resource scheduling method, device, electronic equipment and computer readable storage medium
CN111737003A (en) * 2020-06-24 2020-10-02 重庆紫光华山智安科技有限公司 Pod balanced scheduling method and device, main node and storage medium
US11336588B2 (en) 2020-06-26 2022-05-17 Red Hat, Inc. Metadata driven static determination of controller availability
CN111813556A (en) * 2020-07-21 2020-10-23 北京东方通软件有限公司 Elastic expansion method of virtual cluster in cloud computing environment
CN114327851A (en) * 2020-09-30 2022-04-12 慧与发展有限责任合伙企业 Preventing scheduling or execution of resources on inconsistent host nodes
CN112181661B (en) * 2020-10-13 2023-10-24 极道科技（北京）有限公司 Task scheduling method
CN112181661A (en) * 2020-10-13 2021-01-05 极道科技（北京）有限公司 Task scheduling method
CN112286631A (en) * 2020-10-23 2021-01-29 烽火通信科技股份有限公司 Kubernetes resource scheduling method and device and electronic equipment
CN112363811A (en) * 2020-11-16 2021-02-12 中国电子科技集团公司电子科学研究院 Artificial intelligence computing resource scheduling method and computer readable storage medium
CN112363811B (en) * 2020-11-16 2023-04-07 中国电子科技集团公司电子科学研究院 Artificial intelligence computing resource scheduling method and computer readable storage medium
CN112527454A (en) * 2020-12-04 2021-03-19 上海连尚网络科技有限公司 Container group scheduling method and device, electronic equipment and computer readable medium
US11474905B2 (en) 2020-12-10 2022-10-18 International Business Machines Corporation Identifying harmful containers
CN114625493B (en) * 2020-12-14 2024-04-02 中国石油大学(华东) Kubernetes cluster resource scheduling method based on improved longhorn beetle whisker intelligent method
CN114625493A (en) * 2020-12-14 2022-06-14 中国石油大学(华东) Kubernetes cluster resource scheduling method based on improved intelligent longicorn stigma method
CN112286671A (en) * 2020-12-29 2021-01-29 湖南星河云程信息科技有限公司 Containerization batch processing job scheduling method and device and computer equipment
CN112764906A (en) * 2021-01-26 2021-05-07 浙江工业大学 Cluster resource scheduling method based on user job type and node performance bias
CN112764906B (en) * 2021-01-26 2024-03-15 浙江工业大学 Cluster resource scheduling method based on user job type and node performance bias
CN112988380A (en) * 2021-02-25 2021-06-18 电子科技大学 Kubernetes-based cluster load adjusting method and storage medium
CN113391891A (en) * 2021-05-20 2021-09-14 国网江苏省电力有限公司信息通信分公司 Load balancing resource scheduling method based on Rete and character string pattern matching algorithm
CN113391891B (en) * 2021-05-20 2024-03-12 国网江苏省电力有限公司信息通信分公司 Load balancing resource scheduling method based on Rete and character string pattern matching algorithm
CN113485792B (en) * 2021-07-08 2023-05-26 厦门服云信息科技有限公司 Pod scheduling method in kubernetes cluster, terminal equipment and storage medium
CN113485792A (en) * 2021-07-08 2021-10-08 厦门服云信息科技有限公司 Pod scheduling method in kubernets cluster, terminal equipment and storage medium
US11706105B2 (en) 2021-12-14 2023-07-18 International Business Machines Corporation Selecting low priority pods for guaranteed runs
CN114327886A (en) * 2021-12-24 2022-04-12 国家石油天然气管网集团有限公司 Dynamic resource scheduling method based on big data deep learning
CN114064296A (en) * 2022-01-18 2022-02-18 北京建筑大学 Kubernetes scheduling method, Kubernetes scheduling device and storage medium
CN114064296B (en) * 2022-01-18 2022-04-26 北京建筑大学 Kubernetes scheduling method, Kubernetes scheduling device and storage medium
WO2024021467A1 (en) * 2022-07-26 2024-02-01 天翼云科技有限公司 Cluster resource planning method, device, apparatus, and medium
CN115297112A (en) * 2022-07-31 2022-11-04 南京匡吉信息科技有限公司 Dynamic resource quota and scheduling component based on Kubernetes
CN117112239A (en) * 2023-10-23 2023-11-24 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Extensible load balancing method and system on heterogeneous reasoning back end
CN117112239B (en) * 2023-10-23 2024-02-09 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Extensible load balancing method and system on heterogeneous reasoning back end

Also Published As

Publication number Publication date
CN109960585B (en) 2021-05-14

Similar Documents

Publication Publication Date Title
CN109960585A (en) A kind of resource regulating method based on kubernetes
WO2021208546A1 (en) Multi-dimensional resource scheduling method in kubernetes cluster architecture system
CN110297699B (en) Scheduling method, scheduler, storage medium and system
EP2710470B1 (en) Extensible centralized dynamic resource distribution in a clustered data grid
CN111966500B (en) Resource scheduling method and device, electronic equipment and storage medium
CN110231976B (en) Load prediction-based edge computing platform container deployment method and system
CN109788046B (en) Multi-strategy edge computing resource scheduling method based on improved bee colony algorithm
WO2016082370A1 (en) Distributed node intra-group task scheduling method and system
CN109120715A (en) Dynamic load balancing method under a kind of cloud environment
CN110221920B (en) Deployment method, device, storage medium and system
CN105491150A (en) Load balance processing method based on time sequence and system
CN111147604B (en) Load balancing method for edge calculation of Internet of vehicles
CN111666158A (en) Kubernetes-based container scheduling method and device, storage medium and electronic equipment
CN108427602A (en) A kind of coordinated dispatching method and device of distributed computing task
CN117155942A (en) Micro-service dynamic self-adaptive client load balancing method and system
CN113672391B (en) Parallel computing task scheduling method and system based on Kubernetes
CN114691372A (en) Group intelligent control method of multimedia end edge cloud system
CN111782627B (en) Task and data cooperative scheduling method for wide-area high-performance computing environment
CN110958192B (en) Virtual data center resource allocation system and method based on virtual switch
CN114978913B (en) Cross-domain deployment method and system for service function chains based on cut chains
CN115562841B (en) Cloud video service self-adaptive resource scheduling system and method
CN115686826A (en) Dynamic scheduling method and system based on space-based distributed satellite cluster
WO2021000694A1 (en) Method for deploying services and scheduling apparatus
CN112306642A (en) Workflow scheduling method based on stable matching game theory
CN117215719A (en) Containerized service deployment method based on relative resource demand and load update

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