CN115914225B - Optimization method for Raft consensus algorithm election stage - Google Patents
Optimization method for Raft consensus algorithm election stage Download PDFInfo
- Publication number
- CN115914225B CN115914225B CN202211336170.8A CN202211336170A CN115914225B CN 115914225 B CN115914225 B CN 115914225B CN 202211336170 A CN202211336170 A CN 202211336170A CN 115914225 B CN115914225 B CN 115914225B
- Authority
- CN
- China
- Prior art keywords
- node
- value
- credit
- credit value
- period
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 27
- 238000005457 optimization Methods 0.000 title claims abstract description 8
- 230000004044 response Effects 0.000 claims abstract description 42
- 230000008859 change Effects 0.000 claims abstract description 32
- 230000008569 process Effects 0.000 claims abstract description 24
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000004883 computer application Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 230000010076 replication Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
An optimization method for Raft consensus algorithm election stage comprises the following steps: step 1: obtaining a credit value of each node in the cluster in a current consensus period; step 2: substituting the current credit value of the node into a formula to calculate so as to obtain RET duration of all nodes in the cluster in the current consensus period; step 3: each node randomly distributes a time length within a RET time length range corresponding to each node to overtime, and the node which completes overtime first indexes the tickets to other nodes to obtain more than half of votes, and then the nodes are selected as the master node; step 4: in the consensus process, a master node responds to a message request of a client, and a slave node responds to the message request of the master node and records the response conditions of the master node and the slave node. Step 5: and according to the recorded response condition, substituting the record into a formula to calculate a credit value change value of each node, and updating the current credit value of the node according to the credit value change value.
Description
Technical Field
The invention belongs to the technical field of blockchain, and particularly relates to an optimization method aiming at Raft consensus algorithm election stage.
Background
With the rapid development of blockchain technology and application in various industries, the efficiency requirements for blockchain systems are increasing. The efficiency of the consensus algorithm determines the overall efficiency of the system. Raft as a strong-leadership consensus algorithm, the consensus process can be mainly divided into an election stage and a log replication stage. The broadcasting and forwarding efficiency of the master node influences the consensus efficiency of the cluster, and if the master node is down, the whole cluster cannot work and reenters the election process. The choice of the master node is therefore an important part of the overall process, and the stability of the system is also dependent on the quality of the master node. The conventional Raft consensus algorithm uses a method of random election timeout (Random Election Timeout, hereinafter referred to as RET) to select a master node, that is, all nodes are assigned to random values within a duration range, and simultaneously start a timeout process, a node with a shorter random duration will first complete the timeout process, and ask other nodes for tickets, and when the node obtains more than half of the tickets in a cluster, the node is selected as the master node in the next period.
When the number of nodes in the cluster is large, the Raft algorithm is easy to generate the situation of election conflict in the election stage, namely, random timeout time periods of a plurality of nodes are very close, timeout is almost finished at the same time, and the rest nodes are subjected to vote, finally, each vote-searching node cannot acquire more than half of votes in the cluster, and re-election is needed at the moment. Under a large number of cluster environments, the traditional Raft election mode can result in low election efficiency and repeated communication overhead. Since the application environment of a blockchain typically needs to support a large number of nodes, it is desirable to avoid election conflicts as much as possible and to select a premium node to select as the master node.
Currently, in the research of Raft algorithm, wang Jindong et al propose a grouping consensus algorithm scheme combining Raft and PBFT in the practical Bayesian fault tolerance consensus algorithm based on Raft algorithm improvement published in the journal of computer application at the 7 th period of 2022, the nodes in the group follow Raft consensus based on PageRank optimization, and the main node participates in PBFT consensus followed by the groups as a group leader, and through grouping consensus, the consensus efficiency of the algorithm can be improved on the basis of having Bayesian fault tolerance, and the communication cost is reduced; wu Yusen et al in the "PBFT consensus algorithm improvement scheme design based on grouping Raft mechanism" published in journal of electronic technology and software engineering in 24 th period of 2021, proposed a clustering consensus mechanism based on Raft and K-means clustering algorithm, and in the cluster, raft consensus algorithm with supervising nodes is used, and in the cluster, center nodes participate in PBFT consensus algorithm, so that throughput and consensus efficiency are improved. However, the above methods can not solve the conflict problem of Raft algorithm in the election stage, and can not ensure the quality of the master node and can not improve the stability of the cluster. Regarding how to improve the efficiency of the consensus algorithm on the basis of guaranteeing the quality of the master node and the stability of the cluster, the method provides a Raft consensus algorithm scheme combining with the credit value.
Disclosure of Invention
The invention aims to solve the technical problems that the existing Raft consensus algorithm generates election conflict due to the traditional random overtime election mode, and the cluster stability is poor due to uncontrollable quality of a master node with strong leadership.
An optimization method for Raft consensus algorithm election stage comprises the following steps:
step 1: obtaining a credit value of each node in the cluster in a current consensus period;
Step 2: substituting the current credit value of the node into a formula to calculate so as to obtain RET duration of all nodes in the cluster in the current consensus period;
step 3: each node randomly distributes a time length within a RET time length range corresponding to each node to overtime, and the node which completes overtime first indexes the tickets to other nodes to obtain more than half of votes, and then the nodes are selected as the master node;
step 4: in the consensus process, a master node responds to a message request of a client, and a slave node responds to the message request of the master node and records the response conditions (including response time and downtime times) of the master node and the slave node.
Step 5: and according to the recorded response condition, substituting the record into a formula to calculate a credit value change value of each node, and updating the current credit value of the node according to the credit value change value.
In step 1, the definition of the current credit value of the node meets the following requirements:
1) If the node is a consensus node newly added into the cluster, directly distributing an initial credit value C of the node;
2) If the node is an original node in the cluster, the node means that after the consensus process of the previous period is finished, the updated credit value obtained according to the self-expression, namely the node credit value in the nth period is Wherein credit n represents the credit value of the node in the nth consensus period, and C is the initial credit value; Δcredit is the value of the credit change of the node in the present consensus period, namely the credit change value; the credit value of the current period of the node is the sum of the initial value of the credit value and the credit change value of the first n consensus periods.
In step 2, the formula for calculating the RET duration of the node is:
1) When the node performs well in the consensus process of the previous period, namely, the change value of the credit value is positive number (deltacredit > 0), the upper and lower limit values of the RET range of the node satisfy the formula:
Timemax＝Tmax-Δcredit·N·k (1)
Timemin＝Tmin-Δcredit·N·2k (2)
Wherein Time max is the upper limit of the node RET duration range and Time min is the lower limit of the node RET duration range; t max is the upper limit value of RET duration range of the node in the last consensus period, and T min is the lower limit value of RET duration range of the node in the last consensus period; deltacredit is the node credit change value; n is the total number of nodes in the current cluster system, and k represents the variation parameter of RET duration range;
2) When the node performs worse in the consensus process of the previous period, namely the change value of the credit value is a negative number (delta credit < 0), the upper and lower limit values of the RET range of the node satisfy the formula:
Timemax＝Tmax-Δcredit·N·2k (3)
Timemin＝Tmin-Δcredit·N·k (4)
Wherein Time max is the upper limit of the node RET duration range and Time min is the lower limit of the node RET duration range; t max is the upper limit value of RET duration range of the node in the last consensus period, and T min is the lower limit value of RET duration range of the node in the last consensus period; deltacredit is the node credit change value; n is the total number of nodes in the current cluster system, and k represents the variation parameter of RET duration range.
In step 4, analyzing the response conditions of the master node and the slave node respectively from response time length and downtime times, wherein the response time length refers to time length required by the node to complete an instruction, and the downtime times refer to times of no response of the node to the instruction in the current period;
the response conditions of the master node and the slave node are respectively as follows:
1) The response time of the master node refers to the time duration of sending the information request sent by the client to the rest slave nodes through the heartbeat interval (APPEND ENTRIES);
2) The downtime times of the master node refer to the times that the master node does not respond to the information request sent by the client;
3) The response time of the slave node refers to the time duration for replying to the master node after the slave node receives the instruction of the master node;
4) The downtime times of the slave nodes refer to the times that the slave nodes do not reply to the heartbeat intervals sent by the master nodes.
In step 5, the change values for changing the respective credit values according to the behavior of the master node and the slave node satisfy the following calculation formula:
Wherein r represents the role of the node (master node or slave node), Δcredit represents the variation of the credit value generated by the node after a certain period passes, if Δcredit is greater than 0, it represents the increase of the credit value of the node, and if Δcredit is less than 0, it represents the decrease of the credit value of the node; n represents the number of tenns in a period; e (t r) represents the expected value of the response time of the role node, t i represents the actual value of the response time of the node in the current period, S represents the influence coefficient of the response time on the credit value, and S is a positive value. D represents an influence coefficient of downtime times on a credit value, D is a negative value, and M F represents downtime times of the node in a wilt period; according to the formula (5), when the response speed of the node is higher than the expected speed, the credit value is increased, otherwise the credit value is decreased; the more times the node is down, the more credit value is reduced; and accumulating the node performance conditions in n periods to obtain the credit value variation delta credit in a certain period.
Compared with the prior art, the invention has the following technical effects:
1) The parameters of the credit value formula in the invention can be adjusted according to the total number of nodes in the cluster, and the parameter values of the influence of the node performance on the credit value can be manually set, so that the method has better universality and self-definition capability;
2) The invention defines the credit value of the node through the expression of the node in the consensus process, and controls the probability of selecting the master node according to the credit value, so that the node with good expression in the cluster has higher probability of selecting the master node, thereby improving the quality of the master node and the stability of the cluster.
Drawings
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
FIG. 1 is a flow chart of the present invention;
fig. 2 is a schematic diagram showing comparison of election time delay experiments according to the present invention.
Detailed Description
As shown in fig. 1, an optimization method for Raft consensus algorithm election phase includes the following steps:
step 1: obtaining a credit value of each node in the cluster in a current consensus period;
Step 2: substituting the current credit value of the node into a formula to calculate so as to obtain RET duration of all nodes in the cluster in the current consensus period;
step 3: each node randomly distributes a time length within a RET time length range corresponding to each node to overtime, and the node which completes overtime first indexes the tickets to other nodes to obtain more than half of votes, and then the nodes are selected as the master node;
step 4: in the consensus process, a master node responds to a message request of a client, and a slave node responds to the message request of the master node and records the response conditions (including response time and downtime times) of the master node and the slave node.
Step 5: and according to the recorded response condition, substituting the record into a formula to calculate a credit value change value of each node, and updating the current credit value of the node according to the credit value change value.
The method is characterized in that in the step 1, the definition of the current credit value of the node meets the following requirements:
1) If the node is a consensus node newly added into the cluster, directly distributing an initial credit value C=50;
2) If the node is an original node in the cluster, the node means that after the consensus process of the previous period is finished, the updated credit value obtained according to the self-expression, namely the node credit value in the nth period is Wherein credit n represents the credit value of the node in the nth consensus period, and C is the initial credit value; Δcredit is the value of the credit change of the node in the present consensus period, namely the credit change value; the credit value of the current period of the node is the sum of the initial value of the credit value and the credit change value of the first n consensus periods.
In step 2, the formula for calculating the RET duration of the node is:
1) When the node performs well in the consensus process of the previous period, namely, the change value of the credit value is positive number (deltacredit > 0), the upper and lower limit values of the RET range of the node satisfy the formula:
Timemax＝Tmax-Δcredit·N·k (1)
Timemin＝Tmin-Δcredit·N·2k (2)
Wherein Time max is the upper limit of the node RET duration range and Time min is the lower limit of the node RET duration range; t max is the upper limit value of RET duration range of the node in the last consensus period, and T min is the lower limit value of RET duration range of the node in the last consensus period; deltacredit is the node credit change value; n is the total number of nodes in the current cluster system, k represents the variation parameter of RET duration range, and the value of k is set to be 1;
2) When the node performs worse in the consensus process of the previous period, namely the change value of the credit value is a negative number (delta credit < 0), the upper and lower limit values of the RET range of the node satisfy the formula:
Timemax＝Tmax-Δcredit·N·2k (3)
Timemin＝Tmin-Δcredit·N·k (4)
wherein Time max is the upper limit of the node RET duration range and Time min is the lower limit of the node RET duration range; t max is the upper limit value of RET duration range of the node in the last consensus period, and T min is the lower limit value of RET duration range of the node in the last consensus period; deltacredit is the node credit change value; n is the total number of nodes in the current cluster system, k represents the variation parameter of RET duration range, and the value of k is set to be 1.
In step 4, analyzing the response conditions of the master node and the slave node respectively from response time length and downtime times, wherein the response time length refers to time length required by the node to complete an instruction, and the downtime times refer to times of no response of the node to the instruction in the current period;
the response conditions of the master node and the slave node are respectively as follows:
1) The response time of the master node refers to the time duration of sending the information request sent by the client to the rest slave nodes through the heartbeat interval (APPEND ENTRIES);
2) The downtime times of the master node refer to the times that the master node does not respond to the information request sent by the client;
3) The response time of the slave node refers to the time duration for replying to the master node after the slave node receives the instruction of the master node;
4) The downtime times of the slave nodes refer to the times that the slave nodes do not reply to the heartbeat intervals sent by the master nodes.
In step 5, the change values for changing the respective credit values according to the behavior of the master node and the slave node satisfy the following calculation formula:
Wherein r represents the role of the node (master node or slave node), Δcredit represents the variation of the credit value generated by the node after a certain period passes, if Δcredit is greater than 0, it represents the increase of the credit value of the node, and if Δcredit is less than 0, it represents the decrease of the credit value of the node; n represents the number of tenns in a period; e (t r) represents the expected value of the response time of the role node, t i represents the actual value of the response time of the node in the current period, S represents the influence coefficient of the response time on the credit value, and S is a positive value. D represents an influence coefficient of downtime times on a credit value, D is a negative value, and M F represents downtime times of the node in a wilt period; according to the formula (5), when the response speed of the node is higher than the expected speed, the credit value is increased, otherwise the credit value is decreased; the more times the node is down, the more credit value is reduced; and accumulating the node performance conditions in n periods to obtain the credit value variation delta credit in a certain period.
According to the method, different credit values are given to all nodes in a cluster through a credit value combination mechanism in an election stage of Raft consensus algorithm, the nodes are screened, different timeout periods are distributed to the nodes according to the credit value of the nodes, after calculation, the node with the high credit value obtains a shorter timeout period, and after calculation, the node with the low credit value obtains a longer timeout period. And screening out a part of good-performance high-quality nodes by using a credit value mechanism, so that the high-quality nodes can finish the overtime process more quickly, and the probability of selecting the high-quality nodes as the main nodes is improved from the probability angle. Meanwhile, the overtime time of the node can be randomly generated from different time ranges, the probability that a plurality of nodes are distributed to similar overtime time can be reduced, and further the occurrence of election conflict is effectively avoided.
Examples:
The invention simulates a block chain cluster with 10 nodes by using Goland language in a windows system, applies the method to the cluster, sets the initial credit value of all the nodes as C=50, simulates 20 consensus periods, and updates the credit value of each node according to the corresponding formula after each consensus period is finished.
Before the algorithm starts the election process, all nodes calculate the upper and lower limits of RET to be allocated to each node in the consensus period according to the credit value of the current period. And then, all nodes in the cluster randomly select overtime in the RET range, select a master node, record the performance of the master node and slave nodes in each consensus process, calculate delta credit according to a formula after the current consensus period is finished, and update the credit value of the node as the basis for calculating the RET range in the next consensus period.
After segmenting RET duration ranges of the nodes through a credit value mechanism, comparing the performance of 10 nodes in 20 consensus periods, recording the consensus duration required by completing each period, setting the RET range to 1000-5000 ms through experiments, indicating that election conflict occurs through the consensus process of exceeding 3500ms, comparing the prior Raft consensus in the 20 consensus periods through experiments, wherein the improved consensus algorithm (called SRaft in the experiments) has no election conflict condition, and the consensus efficiency is relatively improved by 35%.
The invention combines credit value mechanism in Raft consensus, and periodically and dynamically updates the credit value according to the performance of the node in the consensus process. And when each new consensus period starts, the RET timeout duration range of the node in the election stage of the period is calculated according to the current credit value, so that the probability of election conflict is reduced, and the efficiency and the cluster stability of the election stage are improved.
Claims (2)
1. An optimization method for Raft consensus algorithm election stage is characterized by comprising the following steps:
step 1: obtaining a credit value of each node in the cluster in a current consensus period;
Step 2: substituting the current credit value of the node into a formula to calculate so as to obtain RET duration of all nodes in the cluster in the current consensus period;
step 3: each node randomly distributes a time length within a RET time length range corresponding to each node to overtime, and the node which completes overtime first indexes the tickets to other nodes to obtain more than half of votes, and then the nodes are selected as the master node;
Step 4: in the consensus process, a master node responds to a message request of a client, a slave node responds to the message request of the master node, and the response conditions of the master node and the slave node are recorded;
Step 5: substituting the recorded response conditions into a formula to calculate a credit value change value of each node, and updating the current credit value of the node according to the credit value change value;
In step 1, the definition of the current credit value of the node meets the following requirements:
1) If the node is a consensus node newly added into the cluster, directly distributing an initial credit value C of the node;
2) If the node is an original node in the cluster, the node means that after the consensus process of the previous period is finished, the updated credit value obtained according to the self-expression, namely the node credit value in the nth period is Wherein credit n represents the credit value of the node in the nth consensus period, and C is the initial credit value; Δcredit is the value of the credit change of the node in the present consensus period, namely the credit change value; the credit value of the current period of the node is the sum of the initial value of the credit value and the credit change value of the first n consensus periods;
In step 2, the formula for calculating the RET duration of the node is:
1) When the node performs well in the consensus process of the previous period, namely, the change value of the credit value is positive number, namely, deltacredit > 0, the RET range upper and lower limit values of the node meet the formulas (1) - (2):
Timemax＝Tmax-Δcredit·N·k (1)
Timemin＝Tmin-Δcredit·N·2k (2)
Wherein Time max is the upper limit of the node RET duration range and Time min is the lower limit of the node RET duration range; t max is the upper limit value of RET duration range of the node in the last consensus period, and T min is the lower limit value of RET duration range of the node in the last consensus period; deltacredit is the node credit change value; n is the total number of nodes in the current cluster system, and k represents the variation parameter of RET duration range;
2) When the node performs worse in the consensus process of the previous period, namely, the change value of the credit value is negative, namely, deltacredit is less than 0, the upper and lower limit values of RET range of the node meet the formulas (3) - (4):
Timemax＝Tmax-Δcredit·N·2k (3)
Timemin＝Tmin-Δcredit·N·k (4)
Wherein Time max is the upper limit of the node RET duration range and Time min is the lower limit of the node RET duration range; t max is the upper limit value of RET duration range of the node in the last consensus period, and T min is the lower limit value of RET duration range of the node in the last consensus period; deltacredit is the node credit change value; n is the total number of nodes in the current cluster system, and k represents the variation parameter of RET duration range;
In step 5, the change values for changing the respective credit values according to the behavior of the master node and the slave node satisfy the following calculation formula:
Wherein r represents the role of the node, Δcredit represents the variation of the credit value generated by the node after a certain period passes, if Δcredit >0, it represents the increase of the credit value of the node, and if Δcredit <0, it represents the decrease of the credit value of the node; n represents the number of tenns in a period; e (t r) represents an expected value of the response time of the role node, t i represents an actual value of the response time of the node in the current period, S represents an influence coefficient of the response time on the credit value, and S is a positive value; d represents an influence coefficient of downtime times on a credit value, D is a negative value, and M F represents downtime times of the node in a wilt period; according to the formula (5), when the response speed of the node is higher than the expected speed, the credit value is increased, otherwise the credit value is decreased; the more times the node is down, the more credit value is reduced; and accumulating the node performance conditions in n periods to obtain the credit value variation delta credit in a certain period.
2. The method according to claim 1, wherein in step 4, the response conditions of the master node and the slave node are analyzed respectively from a response time length and a downtime frequency, wherein the response time length refers to a time length required by the node to complete an instruction, and the downtime frequency refers to a frequency of the node not responding to the instruction in the current period;
the response conditions of the master node and the slave node are respectively as follows:
1) The response time of the master node refers to the time duration of sending the information request sent by the client to the rest slave nodes through the heartbeat interval APPEND ENTRIES;
2) The downtime times of the master node refer to the times that the master node does not respond to the information request sent by the client;
3) The response time of the slave node refers to the time duration for replying to the master node after the slave node receives the instruction of the master node;
4) The downtime times of the slave nodes refer to the times that the slave nodes do not reply to the heartbeat intervals sent by the master nodes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211336170.8A CN115914225B (en) | 2022-10-28 | 2022-10-28 | Optimization method for Raft consensus algorithm election stage |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211336170.8A CN115914225B (en) | 2022-10-28 | 2022-10-28 | Optimization method for Raft consensus algorithm election stage |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115914225A CN115914225A (en) | 2023-04-04 |
| CN115914225B true CN115914225B (en) | 2024-04-30 |
Family
ID=86494156
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211336170.8A Active CN115914225B (en) | 2022-10-28 | 2022-10-28 | Optimization method for Raft consensus algorithm election stage |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115914225B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109660367A (en) * | 2018-11-21 | 2019-04-19 | 语联网（武汉）信息技术有限公司 | Reach method, apparatus and electronic equipment based on the common recognition for improving Raft algorithm |
| WO2019222993A1 (en) * | 2018-05-25 | 2019-11-28 | 北京大学深圳研究生院 | Blockchain consensus method based on trust relationship |
| WO2019232789A1 (en) * | 2018-06-08 | 2019-12-12 | 北京大学深圳研究生院 | Voting-based consensus method |
| CN114490125A (en) * | 2022-01-19 | 2022-05-13 | 山东浪潮科学研究院有限公司 | Optimization method for preselection process in Raft consensus algorithm |
-
2022
- 2022-10-28 CN CN202211336170.8A patent/CN115914225B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019222993A1 (en) * | 2018-05-25 | 2019-11-28 | 北京大学深圳研究生院 | Blockchain consensus method based on trust relationship |
| WO2019232789A1 (en) * | 2018-06-08 | 2019-12-12 | 北京大学深圳研究生院 | Voting-based consensus method |
| CN109660367A (en) * | 2018-11-21 | 2019-04-19 | 语联网（武汉）信息技术有限公司 | Reach method, apparatus and electronic equipment based on the common recognition for improving Raft algorithm |
| CN114490125A (en) * | 2022-01-19 | 2022-05-13 | 山东浪潮科学研究院有限公司 | Optimization method for preselection process in Raft consensus algorithm |
Non-Patent Citations (2)
| Title |
|---|
| The research of the optimized solutions to Raft consensus algorithm based on a weighted PageRank algorithm;YUSEN WU! Et.AL;《2022 Asia Conference on Algorithms, Computing and Machine Learning 》;20220819;784-789 * |
| 改进的Raft一致性算法及其研究;陈陆;黄树成;徐克辉;;江苏科技大学学报(自然科学版);20180815(04);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115914225A (en) | 2023-04-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108810140B (en) | High-performance hierarchical storage optimization method based on dynamic threshold adjustment in cloud storage system | |
| CN113378474B (en) | Contribution-based federated learning client selection method, system and medium | |
| CN115481748A (en) | Federal learning freshness optimization method and system based on digital twin assistance | |
| CN115914225B (en) | Optimization method for Raft consensus algorithm election stage | |
| CN111343006B (en) | CDN peak flow prediction method, device and storage medium | |
| CN114866545B (en) | Semi-asynchronous hierarchical federal learning method and system based on air calculation | |
| CN110162716B (en) | Influence community searching method and system based on community retrieval | |
| CN107239548B (en) | Report processing method based on SQL Server and HIVE | |
| CN114665519A (en) | Wind power plant cluster active power coordination control method based on hierarchical model | |
| CN111782396B (en) | Concurrency elastic control method based on distributed database | |
| CN105653654A (en) | Lucky draw qualification indexing system and method | |
| CN115643008A (en) | PBFT algorithm based on partition clustering | |
| CN110119268B (en) | Workflow optimization method based on artificial intelligence | |
| CN116226086A (en) | Sequential maintenance and verification method for segmented data in data migration process | |
| CN108595666B (en) | Method for dynamically calculating news acquisition service resources and information data processing terminal | |
| CN116151599A (en) | Scheduling data processing method based on deep reinforcement learning | |
| CN115712686A (en) | Incremental data synchronization method and device for solving delayed arrival of data | |
| CN114925313A (en) | Self-adaptive method and system based on distributed link tracking dynamic sampling rate | |
| CN113656372B (en) | Standard index database data mart architecture device and method | |
| CN110322078B (en) | Sector boundary flight flow control method and computer storage medium | |
| CN113283699A (en) | Self-adaptive evaluation method based on key index distribution | |
| CN110135747B (en) | Flow customization method based on neural network | |
| CN112330221A (en) | Job shop scheduling optimization method with sufficient necessary condition neighborhood structure | |
| CN112381409A (en) | Dynamic adjustment method for peak time period | |
| CN113360576A (en) | Power grid mass data real-time processing method and device based on Flink Streaming |
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 |