CN104378442B - A kind of Trace file dumping methods for reducing resource contention - Google Patents
A kind of Trace file dumping methods for reducing resource contention Download PDFInfo
- Publication number
- CN104378442B CN104378442B CN201410697278.9A CN201410697278A CN104378442B CN 104378442 B CN104378442 B CN 104378442B CN 201410697278 A CN201410697278 A CN 201410697278A CN 104378442 B CN104378442 B CN 104378442B
- Authority
- CN
- China
- Prior art keywords
- congestion
- information
- service
- lustre
- network
- 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
Landscapes
- Debugging And Monitoring (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of Trace file dumping methods for reducing resource contention, in based on Infiniband networks and Lustre systems, the dump method performs three service steps, i.e. storage information and collects service S1, Infiniband network congestion detection service S2 and file dumping service S3.This method is combined by the IO service conditions of passive detection Lustre systems with the service condition of active probe Infiniband networks, the congestion situation of calculating network and I/O resource, the technological means of trace file dump is carried out when without congestion, so as to realize that reduce trace file dump occurs with the situation that monitored program resource is competed, after acquisition program pitching pile the speed of service faster, the more accurate technique effect of trace data.
Description
Technical field
The present invention relates to the method for one kind trace file (i.e. Trace files) dump in a group of planes, more particularly, refer to
A kind of trace file dump method for reducing resource contention.More particularly to using Infiniband (translation, infinite bandwidth) net
Under network and the cluster environment of Lustre (translation, gloss file system) system, TAU (Tuning and Analysis
Utilities, translation, Commissioning Analysis instrument) software dump procedure.
Background technology
A group of planes is the set of the multiple stand-alone computers interconnected, and these computers can be unit or multiprocessor system
System, such as PC or work station, each node has the memory of oneself, IO (Input/Output, translation, input and output) equipment
And operating system.A group of planes is to user and should be used to say that a single system, and it can provide the high-performance ring of lower price high efficiency
Border and fast and reliable service etc..
Infiniband networks are server I/O of new generation and group of planes intercommunication industry standard network structure.By linking,
Interchanger, CA (Channel Adaptor, translation, channel adapter) and terminal node composition.Its congestion control uses ECN
(Explicit Congestion Notification, translation, explicit congestion notification) method, i.e., have one specially in heading
Door is used for whether representing network congestion.Shown in Figure 1, the CA that Infiniband networks are connected with terminal node is otherwise known as
HCA (Host Channel Adapter, translation, host's channel adapter), the HCA, which are provided, uses Infiniband networks
Software interface.
Lustre systems are the parallel file storage systems for aiming at group of planes design.Shown in Figure 1, Lustre systems are by one
Individual MDS (Metadata Server, translation, meta data server), several OSS (Object Storage Servers, translation,
Object storage server) composition, wherein, MDS is responsible for filename, file path management, and responds storage request, for request distribution
OSS;OSS is responsible for file content storage, and the unit of storage is band.
TAU is transplantable parallel program performance summary and trace analysis instrument.When monitored program is run, TAU works
Tool will produce trace data, and recorded in file system.In current NOWs, most of data are written into Lustre systems
Unite as in the parallel file of representative.
In recent years, with the growth of group of planes scale, efficient concurrent program increasingly relies on Performance Analysis Software.Performance point
Software is analysed in order to obtain run case information, this information is referred to as trace, and trace is stored in file system and supplies ex-post analysis.Mesh
Before, the storage of data uses parallel file system metadata acquisition tool, most important is Lustre systems, program operation and text more
High performance network, mainly Infiniband networks are shared part dump more.Due to program and the network of data acquisition contention system
And I/O resource, these competitions will influence the transmission speed of trace file dump.
The content of the invention
The purpose of the present invention be for available data capture program with being vied each other system resource by the TAU programs monitored,
Cause program runtime after pitching pile long, measure obtained data is influenceed by resource contention, it is impossible to which full backup is actual
The problem of program operation process.The present invention is led using a kind of to Infiniband networks and Lustre system congestion information
Dynamic, passive detection is combined, and the competition avoiding method of dump is carried out when without congestion.
To achieve the above object, the present invention provides following technical scheme:Based on Infiniband networks and Lustre systems
In system, the Lustre systems include I/O Information Statistics node (4), first the passive collector node of I/O information (1), second
The individual passive collector node of I/O information (2) and the passive collector node of n-th I/O information (3).OSS sections in the Lustre systems
Storage information is run on point and collects service S1, there are IO congestion informationsThe operation storage of MDS nodes
Information collecting service S1, collects describedConstitute collection-congestion information IOAlways;
When the TAU that program performs (5,6,7) in the node trace data file produced need write-in, file dumping service
S3 sends active probe bag requestTMDS nodes to Lustre systems;Then, MDS nodes are by collection-congestion information IOAlways
With packet MDTForm feeds back to Infiniband network congestions detection service S2;Infiniband network congestions detection service
On the one hand S2 extracts collection-congestion information IOAlwaysFile dumping service S3 is transferred to, on the other hand using congestion Rule of judgment R-
F obtains the congestion situation C of current Infiniband networksinf={ a, b }, and by described Cinf={ a, b } is transferred to file
Dump services S3;A represents network congestion, and b represents network not congestion;
File dumping services S3 to the C that receivesinf={ a, b } and IOAlways, judge to be based on using resource judgment condition L-I
The service condition U={ au, bu } of the resource of Infiniband networks and Lustre systems, au represent resource congestion, and bu represents money
Source not congestion, decides whether when caching unfilled, in prior notice TAU unloading trace data file to Lustre systems, with
Avoid possible resource contention.
The advantage that the present invention reduces the trace file dump method of resource contention is:
1. the inventive method can be alleviated due to the scaling concern that resource contention is brought, and improve network and I/O system institute band
The size limit come.
2. the present invention is applied on the NOWs based on Infiniband networks and Lustre system buildings, for reducing
Due to the error that network and I/O resource competition are produced, tracking information is more accurately obtained.
3. the inventive method can speed up dump procedure, reduces invasion of the dump to the program tape that is monitored, improves overall
The speed of service.
Brief description of the drawings
Fig. 1 is the NOWs schematic diagram based on Infiniband networks and Lustre system buildings.
Figure 1A is the IO information transfer schematic diagrames of MDS nodes, OSS nodes.
Fig. 2 is the flow chart of trace file dump of the present invention.
Fig. 2A is congestion Rule of judgment R-F of the present invention flow chart.
Embodiment
Below in conjunction with drawings and examples, the present invention is described in further detail.
It is shown in Figure 1, the NOWs based on Infiniband networks and Lustre system buildings, wherein, it is described
It is passive that Lustre systems include I/O Information Statistics node 4, first the passive collector node 1 of I/O information, second I/O information
Collector node 2 and the passive collector node 3 of n-th I/O information.Include first program in described NOWs and perform node
5th, second program performs node 6 and n-th program performs node 7.Program performs node by running the journey after TAU pitching piles
Sequence, for obtaining program operation Trace files (trace file).
In based on Infiniband networks and Lustre systems, employ execution storage information collect service steps S1,
Infiniband network congestions detect service steps S2 and file dumping service steps S3 to realize the Trace of reduction resource contention
File dumping is handled.
Service S1, production are collected referring to storage information is run shown in Figure 1A, Fig. 2, on the OSS nodes in the Lustre systems
There are IO congestion informationsMDS nodes operation storage information, which is collected, services S1, described in collectionConstitute collection-congestion information IOAlways;
Referring to shown in Figure 1A, Fig. 2, Fig. 2A, when the TAU trace data file produced need write-in, file dumping service
S3 sends active probe bag requestTMDS nodes to Lustre systems;Then, MDS nodes are by collection-congestion information IOAlways
With packet MDTForm feeds back to Infiniband network congestions detection service S2;Infiniband network congestions detection service
On the one hand S2 extracts collection-congestion information IOAlwaysFile dumping service S3 is transferred to, on the other hand using congestion Rule of judgment R-
F obtains the congestion situation C of current Infiniband networksinf={ a, b }, and by described Cinf={ a, b } is transferred to file
Dump services S3;A represents network congestion, and b represents network not congestion;
Referring to shown in Figure 1A, Fig. 2, file dumping services S3 to the C that receivesinf={ a, b } and IOAlways, using resource judgment
Condition L-I judges the service condition U={ au, bu } of the resource based on Infiniband networks and Lustre systems, and au is represented
Resource congestion, bu represents resource not congestion, decides whether when caching unfilled, prior notice TAU unloading trace datas file
Into Lustre systems, to avoid possible resource contention.
In the present invention, collection-congestion information IOAlwaysIn particular content collect service for operation storage information on OSS nodes
S1, there is IO congestion informationsIt is expressed as using aggregate form
The IO congestion informationsInRepresent the congestion of the passive collector node 1 of first I/O information
Information,The congestion information of the passive collector node 2 of second I/O information is represented,Represent that n-th I/O information is passively received
Collect the congestion information of node 3.
In the present invention, packet MDTIn include collection-congestion information IOAlways, packet round-trip delay RTT, packet
Congestion parameter flag, MD is expressed as using aggregate formT={ IOAlways,RTT,flag}。
In the present invention, referring to shown in Fig. 2A, congestion Rule of judgment R-F refers to:First determine whether packet congestion parameter
Whether flag is congestion, and when flag is congestion, then current Infiniband networks are designated as network congestion a;
When flag is not congestion, then judge whether packet round-trip delay is more than or equal to packet round-trip delay threshold value,
If RTT >=RTTThreshold value, then current Infiniband networks be designated as network congestion a;If RTT < RTTThreshold value, then it is current
Infiniband networks are designated as network not congestion b.
In the present invention, resource judgment condition L-I refers to:First to Cinf={ a, b } carries out congestion situation judgement, if Cinf
={ a, b } is network congestion a, then the service condition of the resource based on Infiniband networks and Lustre systems is au;
If Cinf={ a, b } is network not congestion b, then according to the collection of summation-congestion information IOAlways(it is designated as CI/O) judge
The congestion situation of Lustre systems;If CI/O≥CThreshold value, then the use of the resource based on Infiniband networks and Lustre systems
Situation is resource congestion au;If CI/O< CThreshold value, then the service condition of the resource based on Infiniband networks and Lustre systems
For resource not congestion bu.The collection of summation-congestion information CI/ORefer to
A kind of Trace file dumping methods of reduction resource contention of the present invention include following specific service steps:
Storage information collects service steps S1
In the present invention, using passive detection technology, collection and statistics by local information obtain IO service conditions.
S11:Start the service of IO monitoring in operating system on I/O node, passive monitoring is carried out on I/O node.
S12:Passive Test is carried out in OSS nodes, when there are IO application generations every time, its Disk bandwidth utilization rate is recorded, when
Before the quantity and CPU usage, memory usage that are written in parallel to.
S13:Count each nodal information.
The MDS nodes of Lustre systems reclaim the monitoring information of OSS nodes at certain intervals, form the letter of overall utilization rate
Breath.The information of the overall utilization rate will be used as collection-congestion information IOAlwaysRun on Infiniband network congestions detection service S2
In.
Infiniband network congestions detection service steps S2
In the present invention, using active probing technique, detected and wrapped by Infiniband, obtain network congestion.
S21:Local HCA information is collected, Infiniband software interface provides HCA queue length, traffic volume, band
The local informations such as width, data are obtained using the method directly read;
S22:Active probe network congestion is, it is necessary to which the node of dump, I/O data is sent by Infiniband to MDS nodes
Request bag, and record delivery time TS.MDS nodes are received after request, and IO congestion informations are replied at once, are initiated nodes records and are received
Time of reception TR is recorded simultaneously to information.Packet round-trip delay RTT=TR-TS and pre-set threshold value RTTThreshold valueComparing to reflect
Network congestion Cinf={ a, b }.If network congestion a, Infiniband can be set to congestion control flag in bag is replied
Flag=1;If congestion control flag can be set to flag=0 by network not congestion b, Infiniband in bag is replied.Congestion
Control bit is another feature of network congestion.
File dumping service steps S3
In the present invention, global buffer is kept, integrated network, IO congestion situations and caching situation carry out shifting to an earlier date dump.
S31:Dump is serviced.The node of dump is being needed, dump service is run.Service caching good at managing, sends actively
Detection bag, sink information determines dump opportunity.
S32:Start detection opportunity.Detection congestion needs to send request bag, and calls correlation function to obtain system information
Data, expend time and resource.So only it is data cached reach a certain amount of when, can just start detection.Lustre gives tacit consent to
Storage tape size be 1.5M, less than 1.5M can produce once connect and enough data can not be write and fill up a storage tape.Therefore
Cache size is set as 6M, when data reach 1.5M, starts to detect for the first time.Detect at certain intervals repeatedly, it is full until caching,
File is then write in units of 1.5M.
S33:The opportunity of dump.When all resources all be not up to congestion degree, i.e., I/O resource utilization rate below 80%,
TR-TS is less than threshold values and congestion control position is 0, can carry out dump without influenceing other in use, carrying out dump.
S34:The Congestion Avoidance stand-by period.Waiting needs the random time related to number of attempt.The time uses for reference index
Backoff algorithm, minimum timeslice be 0.1 second, total time be timeslice be multiplied with multiple, multiple for number of attempt with it is smaller in 10
Number.Stand-by period is the random number and the product of total time of [0,1].When the stand-by period is 0, write at once.
The present invention is one and run under Infiniband networks and Lustre system environments, it is to avoid data acquisition dump with
The method of monitored program competitive resource.By passively collecting Lustre system IO congestion informations, the Infiniband network equipments
Congestion information, active probe Infiniband networks entirety congestion information, decides whether that dump in advance is cached using these information
File, to avoid resource contention.Program operation information finally can be more accurately obtained, strengthens the scalability of program.
The present invention is made up of two parts, is separately operable on Lustre memory nodes and the node for needing file dumping.
I/O data collection service is run on Lustre memory nodes, the service is responsible for monitoring and records magnetic on Lustre nodes
The wide utilization rate of dribbling, the quantity being written in parallel to and CPU usage, memory usage.Gather seeervice cycle property and obtain data, and
It is back to MDS nodes and supplies subsequent query.
Operating file dump is serviced on the node of dump demand, and the service is responsible for cache management, dump opportunity and determines and turn
Store up file.Cache management includes safeguarding that node level written document is cached, and is that each file sets up a caching, and to need written document
Program provide written document service.While in order to determine dump opportunity, dump service request I/O node information, computing system resource
Race condition.In the case of system is uncontested, in data cached write-in Lustre file system.
The present invention implementation procedure be:After system starts, i.e., run I/O data on all memory nodes of Lustre systems
Collection service, waits the periodic request of data of MDS nodes, collection memory node Disk bandwidth utilization rate, the number being written in parallel to
Amount and CPU usage, memory usage, and resource situation is returned into MDS nodes.MDS nodes are then asked and collect all letters
Breath, and collect for all node hardware service conditions of Lustre systems.The collection service of MDS nodes then waits the IO of dump service
Information request, the ChangeLog that MDS nodes are provided using Lustre systems when there is request reads recent Lustre and uses feelings
Condition, as multidate information, requestor is back to together with hardware service condition.
Data dump in need node, carry out Data Collection before, log-on data dump service.Data dump takes
It is engaged in setting up 3M buffer areas for each file.When there is data to need write-in file, the write functions of routine call dump service,
Its effect is write data into the caching of dump service management.When data are more than 1.5M in caching, dump service is saved to MDS
Point initiates IO information inquiries dialogue, and records the initiation time of dialogue.MDS nodes return to Lustre software and hardware resources and take feelings
Condition, dump service calculates RTT after the information of return is received.Software and hardware information and RTT are compared with default threshold values, can
Whether to obtain various resources in the state for being likely to occur competition.If there is competition, then Random Waiting Time is calculated, calculated
Method is as follows:Minimum timeslice be 0.1 second, total time be timeslice be multiplied with multiple, multiple for number of attempt with it is smaller in 10
Number.Stand-by period is the random number of [0,1] and seizing the opportunity for total time.Tried again after random wait certain time.It is uncontested or slow
File dumping is carried out when being filled with once.Dump least unit is 1.5M, i.e., fixed write-in 1.5M or 3M every time, no matter dump in advance
When caching in size of data, target is to establish the link every time and can all write a storage bar, and the overhead established the link is not wasted.
The present invention is a kind of trace file dump strategy, and to be solved is how to reduce file dumping and monitored program
Produce resource contention technical problem, IO service conditions and active probe that this method passes through passive detection Lustre systems
The service condition of Infiniband networks is combined, the congestion situation of calculating network and I/O resource, and trace is carried out when without congestion
The technological means of file dumping, occurs with the situation that monitored program resource is competed so as to realize and reduce trace file dump, obtains
After calling program pitching pile the speed of service faster, the more accurate technique effect of trace data.
Claims (3)
1. a kind of Trace file dumping methods for reducing resource contention, this method is based on Infiniband networks and Lustre
Run in system, the Lustre systems include I/O Information Statistics node (4), first passive collector node of I/O information
(1), second passive collector node of I/O information (2) and the passive collector node of n-th I/O information (3);It is characterized in that:This turn
Method for storing performs storage information and collects service S1 steps, Infiniband network congestions detection service S2 steps and file dumping clothes
Business S3 steps;
Storage information is run on OSS nodes in the Lustre systems and collects service S1, there are IO congestion informationsMDS nodes operation storage information, which is collected, services S1, described in collectionStructure
Into collection-congestion information IOAlways;The congestion information of first passive collector node of I/O information (1) is represented,Represent the
The congestion information of two passive collector nodes of I/O information (2),Represent gathering around for n-th I/O information passive collector node (3)
Fill in information;
When the TAU trace data file produced need write-in, file dumping service S3 sends active probe bag requestTGive
The MDS nodes of Lustre systems;Then, MDS nodes are by collection-congestion information IOAlwaysWith packet MDTForm feeds back to
Infiniband network congestions detection service S2;On the one hand Infiniband network congestions detection service S2, which extracts, collects-gathers around
Fill in information IOAlwaysFile dumping service S3 is transferred to, current Infiniband is on the other hand obtained using congestion Rule of judgment R-F
The congestion situation C of networkinf={ a, b }, and by described Cinf={ a, b } is transferred to file dumping service S3;A represents that network is gathered around
Plug, b represents network not congestion;
File dumping services S3 to the C that receivesinf={ a, b } and collection-congestion information IOAlways, using resource judgment condition L-I
To judge that the service condition U={ au, bu } of the resource based on Infiniband networks and Lustre systems, au represent that resource is gathered around
Plug, bu represents resource not congestion, decides whether that prior notice TAU unloading trace datas file is arrived when caching unfilled
In Lustre systems, to avoid possible resource contention;
Resource judgment condition L-I refers to:First to Cinf={ a, b } carries out congestion situation judgement, if Cinf={ a, b } gathers around for network
A is filled in, then the service condition of the resource based on Infiniband networks and Lustre systems is au;
If Cinf={ a, b } is network not congestion b, then according to the collection of summation-congestion information CI/OTo judge Lustre systems
Congestion situation;If CI/O≥CThreshold value, then the service condition of the resource based on Infiniband networks and Lustre systems is au;If
CI/O< CThreshold value, then the service condition of the resource based on Infiniband networks and Lustre systems is bu.
2. the Trace file dumping methods according to claim 1 for reducing resource contention, it is characterised in that:Congestion judges
Condition R-F refers to:Whether be congestion, when flag is congestion, then current if first determining whether packet congestion parameter flag
Infiniband networks are designated as network congestion a;
When flag is not congestion, then judge whether packet round-trip delay is more than or equal to packet round-trip delay threshold value, if RTT
≥RTTThreshold value, then current Infiniband networks be designated as network congestion a;If RTT < RTTThreshold value, then current Infiniband nets
Network is designated as network congestion b.
3. the Trace file dumping methods according to claim 1 for reducing resource contention, it is characterised in that:Operate in
On the MDS nodes and OSS nodes of Lustre systems.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410697278.9A CN104378442B (en) | 2014-11-26 | 2014-11-26 | A kind of Trace file dumping methods for reducing resource contention |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410697278.9A CN104378442B (en) | 2014-11-26 | 2014-11-26 | A kind of Trace file dumping methods for reducing resource contention |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104378442A CN104378442A (en) | 2015-02-25 |
CN104378442B true CN104378442B (en) | 2017-08-25 |
Family
ID=52557088
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410697278.9A Active CN104378442B (en) | 2014-11-26 | 2014-11-26 | A kind of Trace file dumping methods for reducing resource contention |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104378442B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US12093134B2 (en) | 2020-04-24 | 2024-09-17 | EMC IP Holding Company LLC | Method, device, and computer storage medium for managing trace record |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109408270B (en) * | 2018-10-18 | 2021-12-03 | 郑州云海信息技术有限公司 | Read-write operation processing method and device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101710339A (en) * | 2009-11-20 | 2010-05-19 | 中国科学院计算技术研究所 | Method and system for controlling data storage in cluster file system and method for creating file |
CN102404201A (en) * | 2011-10-20 | 2012-04-04 | 浪潮电子信息产业股份有限公司 | Method of realizing maximum bandwidth of Lustre concurrent file system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100088437A1 (en) * | 2008-10-06 | 2010-04-08 | Mellanox Technologies Ltd | Infiniband adaptive congestion control adaptive marking rate |
-
2014
- 2014-11-26 CN CN201410697278.9A patent/CN104378442B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101710339A (en) * | 2009-11-20 | 2010-05-19 | 中国科学院计算技术研究所 | Method and system for controlling data storage in cluster file system and method for creating file |
CN102404201A (en) * | 2011-10-20 | 2012-04-04 | 浪潮电子信息产业股份有限公司 | Method of realizing maximum bandwidth of Lustre concurrent file system |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US12093134B2 (en) | 2020-04-24 | 2024-09-17 | EMC IP Holding Company LLC | Method, device, and computer storage medium for managing trace record |
Also Published As
Publication number | Publication date |
---|---|
CN104378442A (en) | 2015-02-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107566206B (en) | Flow measuring method, equipment and system | |
CN102035697B (en) | Concurrent connections performance testing system and method for file system | |
CN101277272B (en) | Method for implementing magnanimity broadcast data warehouse-in | |
EP3070591A1 (en) | Cold and hot data identification threshold calculation method, apparatus and system | |
CN103067297B (en) | A kind of dynamic load balancing method based on resource consumption prediction and device | |
CN105824744A (en) | Real-time log collection and analysis method on basis of B2B (Business to Business) platform | |
CN102063330B (en) | Performance data acquisition method for large-scale parallel program | |
US9792231B1 (en) | Computer system for managing I/O metric information by identifying one or more outliers and comparing set of aggregated I/O metrics | |
CN106959820B (en) | Data extraction method and system | |
CN111562889B (en) | Data processing method, device, system and storage medium | |
CN110147470B (en) | Cross-machine-room data comparison system and method | |
EP2713270A1 (en) | Resource scheduling method and device | |
CN110162270A (en) | Date storage method, memory node and medium based on distributed memory system | |
CN111966289A (en) | Partition optimization method and system based on Kafka cluster | |
CN112260889B (en) | Linux-based process flow monitoring method, system and equipment | |
US20200394904A1 (en) | Metering data management system and computer readable recording medium | |
CN106874319A (en) | The distributed statistical method and device of click volume | |
CN115269108A (en) | Data processing method, device and equipment | |
CN103336730A (en) | Data backup method and system based on cloud storage | |
CN104378442B (en) | A kind of Trace file dumping methods for reducing resource contention | |
US9736215B1 (en) | System and method for correlating end-user experience data and backend-performance data | |
CN106709053A (en) | Data slice management method and system | |
CN103095812A (en) | Copy establishing method based on user request response time | |
CN111953567B (en) | Method, system, equipment and medium for configuring multi-cluster management software parameters | |
CN113422808A (en) | Internet of things platform HTTP information pushing method, system, device and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |