CN107092649A - A kind of topological replacement method of unaware towards real-time stream calculation - Google Patents
A kind of topological replacement method of unaware towards real-time stream calculation Download PDFInfo
- Publication number
- CN107092649A CN107092649A CN201710145598.7A CN201710145598A CN107092649A CN 107092649 A CN107092649 A CN 107092649A CN 201710145598 A CN201710145598 A CN 201710145598A CN 107092649 A CN107092649 A CN 107092649A
- Authority
- CN
- China
- Prior art keywords
- ringbuffer
- topology
- time
- data
- length
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/901—Indexing; Data structures therefor; Storage structures
- G06F16/9024—Graphs; Linked lists
Abstract
A kind of topological replacement method of unaware towards real-time stream calculation, initializes a circle queue from beginning to end；Obtain the quantity failNum of data calculating failure in topology in actual time window；Queue is traveled through, for each element in queue, E (M) is updated；Queue is traveled through, for each element in queue, S is updated2；Judge S2Whether value is more than C, if judging whether faileNum is more than E (M), if carrying out dilatation replacement to current topology more than if, otherwise carries out capacity reducing replacement；Task indexes in current Topology are recalculated；NewTopology is initialized, according to the Worker Node where index distribution Task, waits original Topology data processings to start to perform again after completing, newly arrived data can flow into newTopology.Automatic progress of the invention, system need not be suspended and transparent to user.
Description
Technical field
The present invention relates to areas of information technology, and in particular to a kind of topological replacement side of unaware towards real-time stream calculation
Method.
Background technology
From social networks information (to provide hot issue or in real time search) to advertisement processing data engine, real-time stream calculation
It is widely used in the industry today, such as Apahe Storm, Twitter ' s Heron, Apache Flink, Spark
Streaming, Samza etc..In such systems, the generation of data is determined by data source completely, the dynamic change of data source and
State disunity causes the speed of data flow to present paroxysmal feature, and the bursty nature of data flow was frequently resulted in
The generation of load, occurs overload and also has following reason：Network congestion, resource utilization is high, interference, heterogeneous, the resistance of IO high frequencies
Plug etc..Therefore, in real-time stream calculation, overload is common and is difficult to avoid that.
Real-time stream calculation is applied to big data calculating field by many esbablished corporations, and such as Taobao analyzes in real time, A Liyun
Galaxy is calculated in real time, ctrip.com's station performance monitoring etc..For Real-time System, the response and stability of system are concerns
Emphasis.Response means to reduce the delay of processing data, i.e. data computing relay, for example, data from it input into system to
Its result is reflected to user's elapsed time；Stability means that system stably and lastingly can be run in the cluster.And mistake
The generation of load easily causes the overall data computing relay increase of system and unstable or even unavailable.
In real-time streams computing system, it is a directed acyclic graph (DAG) to calculate structure, is referred to as topological (Topology),
Topology is made up of data flow (Steam), the generator's component (Spout) and operational components (Bolt) of data flow.Task is topology
The example of middle Spout or Bolt operationally, the process for performing Task is referred to as actuator (Executor), the clothes where actuator
Business device is referred to as working node (Worker Node).When real-time streams computing system is overloaded, the number calculated in systems
, can be by distributing more to topological (Topology) to solve this problem according to showing to fail because of calculating time time-out
Computing resource, the concurrency for improving Topology solves.But this solution has limitation, because in Topology
Task quantity once it is determined that can not change, restart topology after business quantity unless repairing is changed to, therefore make in this way
Carry out resource redistributes the task quantity for being also limited to set before system operation.
Both at home and abroad many researchs, the method that J.Xu is proposed have been done on the scheduling of resource of real-time streaming computing system and distribution
Each Executor workload mainly is collected by changing actuator (Executor), and judges the working node
Whether whether (Worker Node) overloads, overloaded to select different scheduling strategies according to Worker Node.L.Aniell is carried
A kind of adaptive on-line scheduling device is gone out, its purpose is also the shortcoming improvement to real-time streams computing system dispatching method, and this two
The core views of person are all to reduce the network traffics between node, but both optimization methods based on scheduling must suspend topology
Carry out resource to redistribute, system is unavailable in re-allocation process, this may result in longer data computing relay and
The loss of data, this is unacceptable for Real-time System.
The content of the invention
Break-Up System is needed to redistributing for computing resource in order to overcome in existing real-time flow calculation methodologies, system is temporary
Disabled during stopping, resource redistributes the deficiency for being limited to preset task quantity, and the present invention proposes a kind of automatic progress
, system without the topological replacement method of pause, transparent to the user unaware towards real-time stream calculation, the method is according to opening up
The overall load flutterred dynamically adjusts PC cluster resource used in topology.
In order to solve the above-mentioned technical problem the present invention provides following technical scheme：
A kind of topological replacement method of unaware towards real-time stream calculation, comprises the following steps：
Step (1) initializes a circle queue ringBuffer from beginning to end, and its length is set to length；
Step (2) obtains the quantity failNum of data calculating failure in topology in actual time window；
Step (3) travels through ringBuffer queues, for each element ringBuffer (i) in ringBuffer,
E (M)=E (M)+ringBuffer (i)；Wherein E (M) initial value is 0, represents data meter in all time windows of ringBuffer
Calculate the average of failure quantity；
Step (4) E (M)=E (M)/A；Wherein A refers to effective TW numbers in current ringBuffer；
Step (5) travels through ringBuffer queues, for each element ringBuffer (i) in ringBuffer,
S2=S2+ (ringBuffer (k)-E (M)) 2*Wt (i), wherein S2For weighting dispersion ratio, Wt (i) is the time window of systemic presupposition
Mouth weights, value is [0.5,1]；
Step (6) S2=S2/A；
Step (7) judges S2Whether value is more than C, if then carrying out step (8), otherwise terminates, without carrying out topological replacement,
C is the dispersion ratio threshold value of systemic presupposition；
Step (8) judges whether faileNum is more than E (M), if carrying out step (9) more than if, otherwise carries out step (10)；
Step (9) carries out dilatation replacement to current topology, and its process is as follows：
9.1) the concurrency currentParallelism of present topology is obtained；
9.2) the copy newTopology of a present topology is replicated in real time computation system, its concurrency is set
For currentParallelism*2, step (11) is carried out；
Step (10) carries out capacity reducing replacement to present topology, and its process is as follows：
10.1) the concurrency currentParallelism of present topology is obtained；
10.2) the copy newTopology of a present topology is replicated in real time computation system, its concurrency is set to
CurrentParallelism/2, carries out step (11)；
Step (11) is recalculated to the Task indexes in current Topology；
Step (12) initializes newTopology, according to the Worker Node where index distribution Task, waits former
Topology data processings start to perform again after completing, and now newly arrived data can flow into newTopology.
Further, in the step (1), the circle queue ringBuffer is the infinite queue of a recycling,
Each element in ringBuffer is a time window TW, and the length of time window is designated as TWL, and TW Data Collections are complete
Into refer to the time window successful collection in time length of window data calculate failure quantity, Data Collection complete TW
Also referred to as effective TW, it follows that the length of the time window of ringBuffer records is length*TWL, ringBuffer meetings
Change over time and ad infinitum elapse backward, the time interval that it is recorded is [current time-length*TWL, current time].
The beneficial effects of the invention are as follows when whole Topology light load, system can reduce Topology automatically
Concurrency (capacity reducing), save PC cluster resource.When whole Topology overloads, system can improve Topology automatically
Concurrency (dilatation), is that Topology distributes more system resources, it is to avoid system is overloaded.The major advantage of this method
It is：1) user is without intervening；2) system is without pause；3) topology dynamically adjusts the computing resource needed according to own load；4)
Resource allocation is not only restricted to the task quantity of systemic presupposition.
Brief description of the drawings
Fig. 1 is average and weighting dispersion ratio calculation formula schematic diagram in the embodiment of the present invention.
Fig. 2 is reducing and expansion appearance index calculation formula schematic diagram in the embodiment of the present invention.
Fig. 3 is reducing and expansion appearance index change schematic diagram in the embodiment of the present invention.
Fig. 4 is unaware topology replacement schematic flow sheet in the embodiment of the present invention.
Fig. 5 is topological (Topology) capacity reducing schematic diagram in the embodiment of the present invention.
Fig. 6 is topological (Topology) dilatation schematic diagram in the embodiment of the present invention.
Embodiment
To make the features described above and process of the present invention more obvious understandable, special embodiment below, and it is detailed to coordinate accompanying drawing to make
Carefully it is described as follows.
Fig. 1 is average and weighting dispersion ratio calculation formula schematic diagram in the embodiment of the present invention.The core that unaware topology is replaced
Thought wants to collect the data bulk for calculating failure in multi-section time window in Topology, then calculates miss data quantity
Average E (M) and weighting dispersion ratio S2, formula is as shown in Figure 1.Wherein Wi represents weights, and weights are related to time window, and distance is worked as
The more long weights of preceding time window are lower, and value is [0.5,1], and the weights of current newest window are 1, apart from farthest window
Weights are 0.5, and other windows are successively decreased with (1-0.5)/length difference, and length is the length of circle queue, value in fact
For 60.S2Value is bigger, represents that current Topology is more unstable.If the quantity that fails in actual time window is more than E (M), and S2
More than max-thresholds C, then dilatation is carried out；If the quantity that fails in actual time window is less than E (M), and S2More than max-thresholds
C.Represent that current Topology is possible to resource excess, carry out capacity reducing.C is the weighting default threshold value of dispersion ratio, and the present embodiment takes
It is worth for 200.
Fig. 2 is that index calculation formula schematic diagram is held in reducing and expansion of the embodiment of the present invention.When topology carries out capacity reducing or dilatation, its
The Worker Node index positions that Task is distributed can change, change formula as shown in Fig. 2 hash is former index value,
Newindex represents the index value after change, and newcap represents the concurrency after Topology change, point following two situations：
(1) Topology carries out dilatation replacement, then newCap value is twice of present topology concurrency (nowcap)；
(2) Topology carries out capacity reducing replacement, then newCap value is the 1/2 of present topology concurrency (nowcap).
Fig. 3 is reducing and expansion appearance index change schematic diagram in the embodiment of the present invention, to further understand Fig. 2 formula, coordinates Fig. 3 to enter
Row explanation.As shown in figure 3, (old) represents that two kinds of key of keyA and keyB before dilatation determine the example of index position, figure in figure
In (new) represent that dilatation latter two key determines the example of index position, wherein Len is Hash table length, hash1, hash2 points
It is not the corresponding Hash of keyA, keyB and high-order operation result.Element is after cryptographic Hash is recalculated, because Hash table length
It is changed into original 2 times, then new index position of the Len-1 scope after high-order 1 bit, therefore change more is as follows：
(1) high-order is 0, then its position is constant after changing, (1) in such as Fig. 3；
(2) high-order is 1, then its position is original 2 times after changing, (2) in such as Fig. 3.
By above-mentioned principle understand have under average case half Task indexes be not required to conversion so that reduce reducing and expansion hold when
Between.
Fig. 4 is unaware topology replacement schematic flow sheet in the embodiment of the present invention.As shown in figure 4, with Apache
Exemplified by the topological replacement method of unaware realized in Storm, it is as follows that it performs step：
Step (1) initializes a circle queue ringBuffer from beginning to end, and its length is set to length..
RingBuffer is the infinite queue of a recycling.Each element in ringBuffer is a time window
(TW), the length of time window be designated as TWL (time window represent a period of time, such as 18:00~18:01 is a time window
Mouthful, then TWL be 1 minute, this window memory storage be topology in the time 18:00~18:Data calculate the number of failure in 01
Amount, TW Data Collections complete to refer to the time window successful collection to 18:00~18:Data calculate the quantity of failure in 01,
The TW that Data Collection is completed is also referred to as effective TW).The length of time window for therefore deducing that ringBuffer records is
length*TWL.RingBuffer can be changed over time ad infinitum to be elapsed backward, it record time interval be [current time-
Length*TWL, current time].So that TWL is 1 minute as an example, length is 60, from 18:00 start recording, then ringBuffer
The time that queue is recorded when expiring for the first time is 18:00~19:00, hereafter earliest time window can be capped, i.e., next minute,
18:00~18:01 time window can be changed into the head of queue, and its time window is changed into 19:00~19:01, ringBuffer note
The time of record is changed into 18:01~19:01；
Step (2) obtains the quantity failNum of data calculating failure in topology in actual time window；
Step (3) travels through ringBuffer queues.For each element ringBuffer (i) in ringBuffer,
E (M)=E (M)+ringBuffer (i)；Wherein E (M) initial value is 0, represents data meter in all time windows of ringBuffer
Calculate the average of failure quantity；
Step (4) E (M)=E (M)/A；Wherein A refers to effective TW numbers in current ringBuffer；
Step (5) travels through ringBuffer queues.For each element ringBuffer (i) in ringBuffer,
S2=S2+(ringBuffer(k)-E(M))2*Wt(i).Wherein S2For weighting dispersion ratio, Wt (i) is the time window of systemic presupposition
Mouth weights, value is [0.5,1], lower apart from the more long time window value of current time；
Step (6) S2=S2/A；
Step (7) judges S2Whether value is more than C, if then carrying out step (8), otherwise terminates, without carrying out topological replacement.
C is the dispersion ratio threshold value of systemic presupposition.
Step (8) judges whether faileNum is more than E (M), carries out step (9), otherwise carries out step (10).
Step (9) carries out dilatation replacement to current topology, and its process is as follows：
9.1) the concurrency currentParallelism of present topology is obtained；
9.2) the copy newTopology of a present topology is replicated in real time computation system, its concurrency is set
For currentParallelism*2；Perform step (11)
Step (10) carries out capacity reducing replacement to present topology, and its process is as follows：
10.1) the concurrency currentParallelism of present topology is obtained；
10.2) the copy newTopology of a present topology is replicated in real time computation system, its concurrency is set to
currentParallelism/2；
Step (11) is recalculated to the Task indexes in current Topology；
Step (12) performs deactive orders to current Topology.Deactive orders are the orders built in Storm,
The order can make the Spout in Topology no longer read new data.
Whether the data in the current Topology of step (13) all calculate completes, if carrying out (14), otherwise continues
Carry out step (13)；
The current Topology of step (14) performs kill orders.Kill orders are the orders built in Storm, and the order can be complete
Full cut-off stops Topology.
Step (15) initializes newTopology, carries out task distribution according to the indexes of Task newly and starts.
Fig. 5 is topological (Topology) capacity reducing schematic diagram in the embodiment of the present invention.As shown in figure 5, as whole Topology
Light load when, to save system resource, Topology concurrency can be reduced, T2 from 4 concurrency capacity reducings to 2 simultaneously
Hair degree, T3 is from 2 concurrency capacity reducings to 1 concurrency.Tuple represents the data of input topology.
Fig. 6 is topological (Topology) dilatation schematic diagram in the embodiment of the present invention.As shown in fig. 6, as whole Topology
During overload, to distribute more resources to Topology, T2 concurrency is from 2 dilatations to 4.Tuple represents the number of input topology
According to.
The present embodiment judges whether Topology stablizes by detecting data calculating failure quantity.Using built in Storm
Aapche Thrift obtain Topology data calculate failure quantity.Apache Thrift are a far call frames
Frame, for obtaining the Topology being currently running performance and relevant information.In order to ensure the order and dependence of data, enter
Row replaces stylish Topology need to again start to perform after old Topology data processings completion.First used in replacement process
The deactive orders that Storm is provided stop Spout operation, reuse the old Topology that kill orders are exited in cluster.
Perform after deactive orders, Storm can nullify the Spout in old Topology, and old Topology message can all be tied quickly
Shu Zhihang, then performs kill orders, Storm can close had been friends in the past Topology Worker, and remove their shape
State, after new Topology redistributes Task, starts to perform.
Claims (2)
1. a kind of topological replacement method of unaware towards real-time stream calculation, it is characterised in that：Comprise the following steps：
Step (1) initializes a circle queue ringBuffer from beginning to end, and its length is set to length；
Step (2) obtains the quantity failNum of data calculating failure in topology in actual time window；
Step (3) travels through ringBuffer queues, for each element ringBuffer (i) in ringBuffer, E (M)
=E (M)+ringBuffer (i)；Wherein E (M) initial value is 0, represents data in all time windows of ringBuffer and calculates mistake
Lose the average of quantity；
Step (4) E (M)=E (M)/A；Wherein A refers to effective TW numbers in current ringBuffer；
Step (5) travels through ringBuffer queues, for each element ringBuffer (i) in ringBuffer, S2=S2
+ (ringBuffer (k)-E (M)) 2*Wt (i), wherein S2For weighting dispersion ratio, Wt (i) weighs for the time window of systemic presupposition
Value, value is [0.5,1]；
Step (6) S2=S2/A；
Step (7) judges S2It is worth and whether is more than C, if then carrying out step (8), otherwise terminate, without carrying out topological replacement, C is to be
Unite default dispersion ratio threshold value；
Step (8) judges whether faileNum is more than E (M), if carrying out step (9) more than if, otherwise carries out step (10)；
Step (9) carries out dilatation replacement to current topology, and its process is as follows：
9.1) the concurrency currentParallelism of present topology is obtained；
9.2) the copy newTopology of a present topology is replicated in real time computation system, its concurrency is set to
CurrentParallelism*2, carries out step (11)；
Step (10) carries out capacity reducing replacement to present topology, and its process is as follows：
10.1) the concurrency currentParallelism of present topology is obtained；
10.2) the copy newTopology of a present topology is replicated in real time computation system, its concurrency is set to
CurrentParallelism/2, carries out step (11)；
Step (11) is recalculated to the Task indexes in current Topology；
Step (12) initializes newTopology, according to the Worker Node where index distribution Task, waits former
Topology data processings start to perform again after completing, and now newly arrived data can flow into newTopology.
2. a kind of topological replacement method of unaware towards real-time stream calculation as claimed in claim 1, it is characterised in that：It is described
In step (1), the circle queue ringBuffer is each in the infinite queue of a recycling, ringBuffer
Element is a time window TW, and the length of time window is designated as TWL, and TW Data Collections complete to refer to time window success
The quantity that data in time window length calculate failure is collected into, the TW that Data Collection is completed is also referred to as effective TW, it follows that
The length of the time window of ringBuffer records is length*TWL, and ringBuffer can be changed over time ad infinitum to pusher
Move, the time interval that it is recorded is [current time-length*TWL, current time].
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710145598.7A CN107092649B (en) | 2017-03-13 | 2017-03-13 | Real-time stream calculation-oriented non-perception topology replacement method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710145598.7A CN107092649B (en) | 2017-03-13 | 2017-03-13 | Real-time stream calculation-oriented non-perception topology replacement method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107092649A true CN107092649A (en) | 2017-08-25 |
| CN107092649B CN107092649B (en) | 2021-04-06 |
Family
ID=59649295
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710145598.7A Active CN107092649B (en) | 2017-03-13 | 2017-03-13 | Real-time stream calculation-oriented non-perception topology replacement method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107092649B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110019398A (en) * | 2017-12-14 | 2019-07-16 | 北京京东尚科信息技术有限公司 | Method and apparatus for output data |
| CN110515680A (en) * | 2018-05-21 | 2019-11-29 | 吕纪竹 | A kind of method of the given delay repeatability of real-time judge big data itself |
| CN110515681A (en) * | 2018-05-21 | 2019-11-29 | 吕纪竹 | A kind of method of the given delay repeatability of real-time judge flow data itself |
| CN113438168A (en) * | 2021-05-21 | 2021-09-24 | 清华大学 | Software and hardware mixed flow table management method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130290554A1 (en) * | 2012-04-26 | 2013-10-31 | Qiming Chen | Open station as a stream analysis operator container |
| CN106034047A (en) * | 2015-04-16 | 2016-10-19 | 腾讯科技（深圳）有限公司 | Data processing method and device |
| CN106095573A (en) * | 2016-06-08 | 2016-11-09 | 北京大学 | The Storm platform operations of a kind of work nest perception divides equally dispatching method |
-
2017
- 2017-03-13 CN CN201710145598.7A patent/CN107092649B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130290554A1 (en) * | 2012-04-26 | 2013-10-31 | Qiming Chen | Open station as a stream analysis operator container |
| CN106034047A (en) * | 2015-04-16 | 2016-10-19 | 腾讯科技（深圳）有限公司 | Data processing method and device |
| CN106095573A (en) * | 2016-06-08 | 2016-11-09 | 北京大学 | The Storm platform operations of a kind of work nest perception divides equally dispatching method |
Non-Patent Citations (2)
| Title |
|---|
| SANGWHAN CHA: "Towards Real-time Streaming Analytics based on Cloud Computing", 《INTERNATIONAL JOURNAL OF BIG DATA》 * |
| 张鹏: "面向大数据的分布式流处理技术综述", 《计算机研究与发展》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110019398A (en) * | 2017-12-14 | 2019-07-16 | 北京京东尚科信息技术有限公司 | Method and apparatus for output data |
| CN110019398B (en) * | 2017-12-14 | 2022-12-02 | 北京京东尚科信息技术有限公司 | Method and apparatus for outputting data |
| CN110515680A (en) * | 2018-05-21 | 2019-11-29 | 吕纪竹 | A kind of method of the given delay repeatability of real-time judge big data itself |
| CN110515681A (en) * | 2018-05-21 | 2019-11-29 | 吕纪竹 | A kind of method of the given delay repeatability of real-time judge flow data itself |
| CN110515681B (en) * | 2018-05-21 | 2024-04-09 | 吕纪竹 | Method for judging given delay repeatability of stream data in real time |
| CN110515680B (en) * | 2018-05-21 | 2024-04-09 | 吕纪竹 | Method for judging given delay repeatability of big data in real time |
| CN113438168A (en) * | 2021-05-21 | 2021-09-24 | 清华大学 | Software and hardware mixed flow table management method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107092649B (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106549806B (en) | A kind of network slice manager and its management method | |
| CN107092649A (en) | A kind of topological replacement method of unaware towards real-time stream calculation | |
| TWI725744B (en) | Method for establishing system resource prediction and resource management model through multi-layer correlations | |
| CN102104496B (en) | Fault tolerance optimizing method of intermediate data in cloud computing environment | |
| CN104572307A (en) | Method for flexibly scheduling virtual resources | |
| CN105868004B (en) | Scheduling method and scheduling device of service system based on cloud computing | |
| WO2021093365A1 (en) | Gpu video memory management control method and related device | |
| CN110647392A (en) | Intelligent elastic expansion method based on container cluster | |
| CN101227416A (en) | Method for distributing link bandwidth in communication network | |
| CN109460339A (en) | The streaming computing system of log | |
| CN104092620A (en) | Method and device for achieving adjustment of network bandwidth | |
| CN105827678B (en) | Communication means and node under a kind of framework based on High Availabitity | |
| CN106375471A (en) | Edge node determination method and device | |
| CN106325984A (en) | Big data task scheduling apparatus | |
| CN106250566A (en) | A kind of distributed data base and the management method of data operation thereof | |
| US11934665B2 (en) | Systems and methods for ephemeral storage snapshotting | |
| Hanafy et al. | A new infrastructure elasticity control algorithm for containerized cloud | |
| CN107395735B (en) | Delay and capacity reduction scheduling method and system for container cluster | |
| CN105610899A (en) | Text file parallel uploading method and device | |
| CN110069457A (en) | One kind is based on scratch file delet method and system in distributed file storage system | |
| CN103338240A (en) | Cloud server automatic monitoring system and method used for monitoring automatic drifting | |
| CN110069565A (en) | A kind of method and device of distributed data base batch data processing | |
| Zhang et al. | A Multi-Agent based load balancing framework in Cloud Environment | |
| CN101833585A (en) | Database server operation control system, method and device | |
| US20230385159A1 (en) | Systems and methods for preventing data loss |
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 |