CN109189547A - The analogy method of Web service under a kind of evolution scene - Google Patents
The analogy method of Web service under a kind of evolution scene Download PDFInfo
- Publication number
- CN109189547A CN109189547A CN201810864577.5A CN201810864577A CN109189547A CN 109189547 A CN109189547 A CN 109189547A CN 201810864577 A CN201810864577 A CN 201810864577A CN 109189547 A CN109189547 A CN 109189547A
- Authority
- CN
- China
- Prior art keywords
- user
- model
- web service
- web
- load
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45504—Abstract machines for programme code execution, e.g. Java virtual machine [JVM], interpreters, emulators
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The present invention relates to a kind of analogy methods of Web service under evolution scene, comprising the following steps: 1) carries out system modelling using lightweight load module to web system and its load;2) determinant attribute of system model is obtained by integrated automation method for digging in user access logs;3) model verifying is carried out to model and is predicted under evolution scene.Compared with prior art, the present invention has many advantages, such as lightweight load module, shortens the simulation run time, is suitable for a variety of evolution scenes.
Description
Technical field
The present invention relates to computer application technologies, more particularly, to a kind of simulation side of Web service under evolution scene
Method.
Background technique
Existing web system becomes to become increasingly complex, but also develops constantly.For maintenance system and reply system
Continuous evolution, administrator usually requires in face of such problems: if the growth of load or system update, to systematicness
Does which type of can have with the service response time of user's perception influence? before any variation, system is executed and assumes analysis
(What-if analysis) is extremely important and necessary.Especially in cloud computing era, since cloud service can give user
Flexible resource provisioning and price are provided, so the frequent trustship of web system is on cloud.Cloud service is system maintenance and cost control
Bring many benefits.In general, the demand of cloud resource may change with the fluctuation that application program loads, it is therefore, pre- in advance
The changes in demand for surveying this resource is to ensure that the key of response time and save the cost.
In most cases, it is assumed that analysis not only needs to design the system model that can describe system architecture, it is also necessary to energy
The load module of real user behavior is enough described.For resource management and capacity planning, load analysis is to directly affect analysis knot
The key foundation of fruit accuracy.But, some researchers have assumed number of users and typical load model in advance, or false
If user conversation arrival meets Poisson process (poisson process), a Prediction program is reruned then to obtain analysis
As a result.This prediction based on non-real real data cannot must accurately reflect real system situation, in the ideal situation, it is assumed that
Problem analysis should be answered based on the journal file of real system, mutually be applied to the negative of resource requirement and performance prediction to obtain
Carry model.Many researchers use the reference applications journey such as TPC-W, RUBiS and SPEC jEnterprise under controlled conditions
The method that sequence is tested to verify them, but most of researchers have ignored influence of the client-cache to resource consumption.
And most of existing load modules are all based on probability graph model.But if in many of graph model node, side
With possible circulation, it is extremely complex to allow for model.Compared to the reference application program of those non-genuine systems, true application is logical
Often with having more complicated page structure and different response file sizes.These all further increase the complexity of load modeling
Property, therefore, in order to cope with the application in real-life, then need a kind of more simple and telescopic method.
In existing what-if analysis methodology another problem that exist is that most research concentrate on it is one or two kinds of main
In computing resource (such as CPU, memory, disk read-write), without regard to bandwidth resources.Since web system is typically based on net
The network-intensive system of network, therefore, bandwidth often become bottleneck, and bandwidth is also a kind of relatively expensive money in cloud market
Source.In reality, public cloud provider such as Amazon and Ali's cloud limit single void usually using the pricing strategy of stagewise
Bandwidth consumption on quasi- machine.In order to manage limited resource, the bandwidth consumption situation for grasping application-specific in time is very
Important, especially possess the large-scale web system of distribution of numerous virtual machines.Currently, the ultimate challenge of prediction bandwidth consumption is
Caching factor, multiplicity and the service response of complexity, the network transmission of the web services of evolution and complexity (such as congestion control and again
It passes).Forecasting problem for bandwidth consumption and it is related to the response time forecasting problem of bandwidth currently without mature method.It is existing
The method based on model cannot all solve these problems, therefore, for web system, it is believed that need a kind of special net
Network modeling and simulation method is analyzed to carry out the hypothesis in terms of bandwidth.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide Web under a kind of evolution scene
The analogy method of service.
The purpose of the present invention can be achieved through the following technical solutions:
The analogy method of Web service under a kind of evolution scene, comprising the following steps:
1) system modelling is carried out using lightweight load module to web system and its load;
2) determinant attribute of system model is obtained by integrated automation method for digging in user access logs;
3) model verifying is carried out to model and is predicted under evolution scene.
In the step 1), lightweight load module is specifically defined as a four-tuple (T, B, WI, BM), wherein T is
The time span of watch window, B are the set of the behavior model of every class user, and WI is intensity of load, and BM is user behavior combination.
The time span of the watch window is the set time length of a setting, and the behavior model includes multiple
Web service request path, the intensity of load is the total quantity of the different user occurred in a watch window, described
User behavior group is combined into the different number combination of every class user.
The web service request path definition is (sc, sto, d, at), wherein sc is the member in service type set G
Element, sto are Start Time Offset, i.e., the time that service request first appears in a watch window, d is
Duration, i.e. the request duration of a certain request, at Inter-arrivalTime, i.e., one in a watch window
Time interval in a watch window between two neighboring request arrival.
In the step 1), system modelling specifically:
System model with multi-layer structure is constructed using OPNET emulation tool, by respectively include physical layer, network layer and
The client and server end of application layer 3-tier architecture is constituted, wherein
On a physical layer, server-side is made of interchanger and server, and client is indicated not using multiple local area network models
Same user group, each local pessimistic concurrency control are connected with each other by IP cloud and gateway model;
In network layer, the browser of client is configured using HTTP version1.1, and system depending on the user's operation
Select TCP setting, TCP setting is consistent with real system setting in the server of server end, bandwidth be limited in server and
It is set in connection attribute between client;
On the application layer, different web servers and file are distributed in using the HTTP application model expression of OPNET Plays
Web services on server, each web services include two specified underlying attributes, respectively Inter-arrival Time
It is the complex attribute table of a service response characteristic with Page Property, the Page Property, to specified every
The quantity and size of embedded object in a kind of service type, and user behavior is described using the Profile model in OPNET,
Each user class possesses the configuration file of specified Start time an offset and Duration, the work station of client
Quantity setting is combined unanimously with user behavior.
The step 2) specifically includes the following steps:
21) web service request is grouped: by text editing distance as the similarity between web service request, and will be similar
Two URI character strings that degree is greater than threshold value are classified as in same sc class;
22) user is clustered using X-means algorithm: using user interest degree as the input of X-means algorithm, institute
The user interest degree I statedSjCalculating formula are as follows:
Wherein, NSjFor the number of request of the service class Sj of user's access, m is the total quantity of service type;
23) property distribution is estimated using Probability Distribution Fitting: is used as according to the common distribution for generating random attribute value candidate
Distribution is matched with the attribute probability distribution calculated in user access logs, and through the test of fitness of fot in candidate's distribution
Select optimal distributed model, when there is no candidate distribution to meet, then using customized probability-distribution function determine attribute with
Machine value.
In the step 3),
Model verifying specifically: assuming that whether can truly reflect instruction by simulating, verifying load module before analysis
Practice the behavior of user in data set;
Prediction specifically: system model is established by completely modeling and excavation on the basis of training data, according to system mould
The various influences generated for web system performance of developing of type simulation and prediction.
The evolution scene are as follows:
When load changes or system changes, according to the key of variation the modification system and load module of setting
Attribute value, wherein load variation includes the variation of variation and the user behavior combination of intensity of load, and system change includes system body
Architecture variation and system service update.
Compared with prior art, the invention has the following advantages that
The invention proposes a simplified lightweight load modules, it has less model attributes, and does not need
The changeability of analog service request circulation and intensity of load, while the runing time of emulation can also be shortened;Propose a kind of collection
At log automatic excavating method, rapidly can obtain load and system property from a large amount of journal files;In evolution field
Under scape, bandwidth consumption and response time are predicted using based on the what-if analysis methodology of network simulation, administrator can be helped to use
Bandwidth Management is effectively performed.
Detailed description of the invention
Fig. 1 is lightweight load module Establishing process figure.
Fig. 2 is to increase loaded simulation and prediction result, wherein figure (2a) is bandwidth consumption with number of users situation of change, figure
(2b) is average response time with number of users situation of change.
Fig. 3 is the prediction result for reducing bandwidth, wherein figure (3a) is the response time in log with network throughput size
Variation, figure (3b) are the response time in simulation result with the variation of network throughput size.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
Embodiment
The embodiment of the present invention is implemented on a popular mobile applications, it is a typical network-intensive
Web system, it is deployed on the dozens of distributed virtual machine in public cloud, is equipped with load balancer and multilayer server set
Group.This example provides a kind of analogy method of Web service under evolution scene to predict bandwidth consumption and response time, such as Fig. 1
It is shown, specifically includes the following steps:
Step 1 establishes a system model with multi-layer structure using OPNET.
OPNET is the network equipment, and communication link, local area network and network cloud provide a large-scale model library.Create net
The physical topology of network intensive equipment system, the present invention identify required node using browsing object palette or using research tool
And link model.Once finding model, it need to be only dragged and dropped into the workspace of project editor, then select interested chain
Model is connect to add the link between node.Different user groups is indicated using several LAN models in this example, in server
End, a multi-tier Web system are made of an interchanger and 4 servers (virtual machine in cloud data center).They pass through
IP cloud is connected with gateway model.Server model has can be with the hardware attributes of detailed configuration, operating system attribute, CPU model
With some unessential attributes can also can be left default setting with the TCP parameter of detailed configuration.In client, specify every
The sum of work station carrys out analog subscriber in a Emulated LAN, and forms certain user behavior combination.The core of system modelling
Problem is configuration Application application model and Profile model (user behavior).Application model is specified to provide web services
Component software attribute, Profile model description application will be used.
Step 2, load module are established.
According to the accessing characteristic of mobile phone user, preset time length of window is 3 minutes;Utilize the URI based on density
Similarity is greater than the URI character string that threshold value eps is 0.57 and returned into same SC by clustering algorithm, and the present embodiment has found 19 groups altogether
Class is serviced, long-tail phenomenon is very universal in web system, i.e., a small number of service types consume most bandwidth, and final we select
Bandwidth consumption maximum preceding 5 class enters user's sorting procedure, and the total bytes of these five types of SC account for the 95% of whole byte numbers
More than.Each user is calculated to the interest measure of this 5 service types, then user's cluster is carried out using X-means, finally obtains
Obtain 6 class of subscribers.In order to keep load module simpler, in every a kind of user, we remove some total bytes and are less than
5% service type, and ensuring that remaining service type always sends byte in each cluster is more than 95%, final result such as table
Shown in 1.
1 user's cluster result of table
Class of subscriber | Main service type | S1 (Bytes%) | S2 (Bytes%) | S3 (Bytes%) | S4 (Bytes%) | S5 (Bytes%) |
C0 | S1 | 97.99 | 0.12 | 0.21 | 0.36 | 0.04 |
C1 | S1,S4,S3 | 60.24 | 3.47 | 7.45 | 26.65 | 1.31 |
C2 | S5 | 0.01 | 0.14 | 1.36 | 0.00 | 97.13 |
C3 | S4,S5 | 2.08 | 0.09 | 3.10 | 89.21 | 3.92 |
C4 | S2,S3,S5 | 3.17 | 59.89 | 24.76 | 2.35 | 9.10 |
C5 | S3,S2,S5 | 1.14 | 5.79 | 84.81 | 0.89 | 6.21 |
Step 3, the excavation of attribute.
The probability distribution of the attribute calculated from log is fitted in this example using candidate distribution, and is tested by goodness
To select best model.Selection criteria is: the Kolmogorov-Smirnov statistic of candidate's distribution is examined less than 0.1 in card side
It is top ranked in testing.This inspection can be rapidly completed by the tool of entitled EasyFit.If without being suitably distributed satisfaction
This selection criteria, we are exactly the customized PDF that the attribute is calculated with program, and cardinal principle is to divide attribute value range into
1000 equal portions, even if the probability of occurrence of each equal portions, according to the probability of each equal portions with this etc. when needing to generate attribute random value
The average generation of part.
Step 4, model verifying.
In order to verify the model, carried out using corresponding statistical indicator calculated in log as reference data and simulation result
Compare, for most frequent 95 user of intensity of load occurs, the correctness of user's number of request emulation is higher than bandwidth consumption
As a result, even if the error of partial service classification is higher than 10%, but global error is respectively less than 10%, in general, model verifying knot
Fruit error can be predicted less than 10%.
2 model verification result of table
Step 5, the prediction under evolution scene.
Scene one: increase load, by taking a file server (virtual machine) as an example, the load for being stepped up the server is strong
Degree, the variation of bandwidth and response time consumed by network simulation prediction file download, as a result as shown in Fig. 2, Fig. 2 a is
Bandwidth consumption is with the variation of number of users, and when number increases to 80 people or more, the practical byte number that sends just is not further added by, and shows band
Width has arrived limits value, and the Microsoft Loopback Adapter on virtual machine keeps to increase and sends byte number, just simulates real scene
Virtual machine network transmission mechanism in lower cloud, Fig. 2 b be average response time with number of users variation, it can be seen that simulation result with
The trend that true log is recorded is consistent.
Scene two: changing band width configuration, such as reduces the supply of some web server (virtual machine) bandwidth, and 5M bandwidth is reduced
For 3M, it is assumed that the bandwidth consumption of system is constant, and the variation prediction result of response time is as shown in figure 3, Fig. 3 a is that user journal exists
The comparison of bandwidth modification front and back, Fig. 3 b are the prediction result in emulation, show that number of users is in 250 people or so after bandwidth is reduced
When will obviously increase, this is similar with actual conditions.
To sum up, the method for the invention can effectively and accurately predict network bandwidth consumption under evolution scene, though not
It can accurately predict the response time of each service request, but it is expected that the variation tendency of average service response time.
Claims (8)
1. the analogy method of Web service under a kind of evolution scene, which comprises the following steps:
1) system modelling is carried out using lightweight load module to web system and its load;
2) determinant attribute of system model is obtained by integrated automation method for digging in user access logs;
3) model verifying is carried out to model and is predicted under evolution scene.
2. the analogy method of Web service under a kind of evolution scene according to claim 1, which is characterized in that the step
It is rapid 1) in, lightweight load module is specifically defined as a four-tuple (T, B, WI, BM), wherein T be watch window time it is long
Degree, B are the set of the behavior model of every class user, and WI is intensity of load, and BM is user behavior combination.
3. the analogy method of Web service under a kind of evolution scene according to claim 2, which is characterized in that the sight
The time span for examining window is the set time length of a setting, and the behavior model includes multiple web service request paths,
The intensity of load is the total quantity of the different user occurred in a watch window, and the user behavior group is combined into often
The different number of class user combines.
4. the analogy method of Web service under a kind of evolution scene according to claim 3, which is characterized in that the web
Service request path definition is (sc, sto, d, at), wherein sc is the element in service type set G, sto Start
Time Offset, i.e., the time that service request first appears in a watch window, d Duration, i.e. an observation window
The request duration of a certain request in mouthful, at is Inter-arrival Time, i.e., two neighboring in a watch window
Time interval between request arrival.
5. the analogy method of Web service under a kind of evolution scene according to claim 4, which is characterized in that the step
It is rapid 1) in, system modelling specifically:
System model with multi-layer structure is constructed using OPNET emulation tool, by respectively including physical layer, network layer and application
The client and server end of layer 3-tier architecture is constituted, wherein
On a physical layer, server-side is made of interchanger and server, and client indicates different using multiple local area network models
User group, each local pessimistic concurrency control are connected with each other by IP cloud and gateway model;
In network layer, the browser of client is configured using HTTP version1.1, and Systematic selection depending on the user's operation
TCP is arranged, and TCP setting is consistent with real system setting in the server of server end, and bandwidth is limited in server and client
It is set in connection attribute between end;
On the application layer, different web servers and file service are distributed in using the HTTP application model expression of OPNET Plays
Web services on device, each web services include two specified underlying attributes, respectively Inter-arrival Time and
Page Property, the Page Property are the complex attribute table of a service response characteristic, each to specify
The quantity and size of embedded object in kind service type, and user behavior is described using the Profile model in OPNET, often
One user class possesses the configuration file of specified Start time an offset and Duration, the work station number of client
Amount setting is combined unanimously with user behavior.
6. the analogy method of Web service under a kind of evolution scene according to claim 1, which is characterized in that the step
It is rapid 2) specifically includes the following steps:
21) web service request is grouped: by text editing distance as the similarity between web service request, and similarity is big
It is classified as in same sc class in two URI character strings of threshold value;
22) user is clustered using X-means algorithm: described using user interest degree as the input of X-means algorithm
User interest degree ISjCalculating formula are as follows:
Wherein, NSjFor the number of request of the service class Sj of user's access, m is the total quantity of service type;
23) property distribution is estimated using Probability Distribution Fitting: candidate distribution is used as according to the common distribution for generating random attribute value
It is matched with the attribute probability distribution calculated in user access logs, and is selected in candidate's distribution by the test of fitness of fot
Optimal distributed model then determines the random value of attribute when not having candidate distribution to meet using customized probability-distribution function.
7. the analogy method of Web service under a kind of evolution scene according to claim 1, which is characterized in that the step
It is rapid 3) in,
Model verifying specifically: assuming that whether can truly reflect trained number by simulating, verifying load module before analysis
According to the behavior for concentrating user;
Prediction specifically: system model is established by completely modeling and excavation on the basis of training data, it is imitative according to system model
Really predict the various influences developed and generated for web system performance.
8. the analogy method of Web service under a kind of evolution scene according to claim 7, which is characterized in that described drills
Change scene are as follows:
When load changes or system changes, according to the determinant attribute of variation the modification system and load module of setting
Value, wherein load variation includes the variation of variation and the user behavior combination of intensity of load, and system change includes system of systems knot
Structure variation and system service update.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810864577.5A CN109189547A (en) | 2018-08-01 | 2018-08-01 | The analogy method of Web service under a kind of evolution scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810864577.5A CN109189547A (en) | 2018-08-01 | 2018-08-01 | The analogy method of Web service under a kind of evolution scene |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109189547A true CN109189547A (en) | 2019-01-11 |
Family
ID=64920288
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810864577.5A Pending CN109189547A (en) | 2018-08-01 | 2018-08-01 | The analogy method of Web service under a kind of evolution scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109189547A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006133820A (en) * | 2004-11-02 | 2006-05-25 | Fujitsu Ltd | Processing time calculation program and workload model creation program |
CN101882105A (en) * | 2010-06-01 | 2010-11-10 | 华南理工大学 | Method for testing response time of Web page under concurrent environment |
CN101916321A (en) * | 2010-09-07 | 2010-12-15 | 中国科学院软件研究所 | Web application fine-grained performance modelling method and system thereof |
CN103383655A (en) * | 2012-01-13 | 2013-11-06 | 埃森哲环球服务有限公司 | Performance interference model for managing consolidated workloads in qos-aware clouds |
-
2018
- 2018-08-01 CN CN201810864577.5A patent/CN109189547A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006133820A (en) * | 2004-11-02 | 2006-05-25 | Fujitsu Ltd | Processing time calculation program and workload model creation program |
CN101882105A (en) * | 2010-06-01 | 2010-11-10 | 华南理工大学 | Method for testing response time of Web page under concurrent environment |
CN101916321A (en) * | 2010-09-07 | 2010-12-15 | 中国科学院软件研究所 | Web application fine-grained performance modelling method and system thereof |
CN103383655A (en) * | 2012-01-13 | 2013-11-06 | 埃森哲环球服务有限公司 | Performance interference model for managing consolidated workloads in qos-aware clouds |
Non-Patent Citations (2)
Title |
---|
JIANPENG HU等: "Log2Sim: Automating What-If Modeling and Prediction for Bandwidth Management of Cloud Hosted Web Services", 《2018 IEEE INTERNATIONAL CONFERENCE ON WEB SERVICES (ICWS)》 * |
YASAMAN AMANNEJAD 等: "Predicting Web service response time percentiles", 《2016 12TH INTERNATIONAL CONFERENCE ON NETWORK AND SERVICE MANAGEMENT (CNSM)》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN100391159C (en) | Method and apparatus for automatic modeling building using inference for IT systems | |
US7739099B2 (en) | Method and system for on-line performance modeling using inference for real production IT systems | |
Verma et al. | Play it again, simmr! | |
US7769843B2 (en) | Apparatus and method for capacity planning for data center server consolidation and workload reassignment | |
US10833937B2 (en) | Detect impact of network maintenance in software defined infrastructure | |
CN113010393A (en) | Fault drilling method and device based on chaotic engineering | |
US8121031B2 (en) | Structure for a multi-scale network traffic generator | |
CN106033469A (en) | A method and system for efficient performance prediction of structured query for big data | |
CN110309052A (en) | A kind of data verification method and relevant device | |
CN105471647B (en) | A kind of power communication network fault positioning method | |
US20060212568A1 (en) | System and method for managing a computer network | |
US20210382775A1 (en) | Systems and methods for classifying and predicting the cause of information technology incidents using machine learning | |
Rak | Response time analysis of distributed web systems using QPNs | |
US20050278301A1 (en) | System and method for determining an optimized process configuration | |
Rak | Performance modeling using queueing Petri nets | |
CN114661463A (en) | BP neural network-based system resource prediction method and system | |
CN113360353A (en) | Test server and cloud platform | |
CN113761193A (en) | Log classification method and device, computer equipment and storage medium | |
CN111988156B (en) | Method for creating network simulation platform, network simulation method and corresponding device | |
CN117170848A (en) | Resource scheduling method and device | |
CN109189547A (en) | The analogy method of Web service under a kind of evolution scene | |
Xie et al. | Modeling traffic of big data platform for large scale datacenter networks | |
Liu et al. | Towards dynamic reconfiguration of composite services via failure estimation of general and domain quality of services | |
Liu et al. | A discrete-event-based simulator for distributed deep learning | |
CN1369983A (en) | Method for simulating WDM optical network |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190111 |
|
WD01 | Invention patent application deemed withdrawn after publication |