CN110188022A - A kind of clustering performance test method and system based on Java - Google Patents
A kind of clustering performance test method and system based on Java Download PDFInfo
- Publication number
- CN110188022A CN110188022A CN201910385999.9A CN201910385999A CN110188022A CN 110188022 A CN110188022 A CN 110188022A CN 201910385999 A CN201910385999 A CN 201910385999A CN 110188022 A CN110188022 A CN 110188022A
- Authority
- CN
- China
- Prior art keywords
- performance test
- cluster
- calculating task
- task
- test
- 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
- 238000011056 performance test Methods 0.000 title claims abstract description 113
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012360 testing method Methods 0.000 claims abstract description 96
- 238000004088 simulation Methods 0.000 claims abstract description 67
- 238000004422 calculation algorithm Methods 0.000 claims description 30
- 238000001629 sign test Methods 0.000 claims description 11
- 238000005516 engineering process Methods 0.000 abstract description 7
- 230000001133 acceleration Effects 0.000 abstract description 2
- 238000011161 development Methods 0.000 abstract description 2
- 238000004891 communication Methods 0.000 description 9
- 230000005540 biological transmission Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000000977 initiatory effect Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000005538 encapsulation Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- ILMNSCQOSGKTNZ-UHFFFAOYSA-N 2-[1-(6-aminopurin-9-yl)-2-oxoethoxy]-3-hydroxypropanal Chemical compound NC1=NC=NC2=C1N=CN2C(OC(CO)C=O)C=O ILMNSCQOSGKTNZ-UHFFFAOYSA-N 0.000 description 1
- 241001269238 Data Species 0.000 description 1
- 238000001545 Page's trend test Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000008140 language development Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3466—Performance evaluation by tracing or monitoring
- G06F11/3476—Data logging
-
- 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
- G06F9/451—Execution arrangements for user interfaces
Abstract
The invention belongs to field of computer technology, in particular to a kind of clustering performance test method and system based on Java, the performance test methods include: to be instructed using JSP simulation test page creation performance test；It is instructed according to the performance test, generates batch calculating task, JNI front-end proxy agent interface is called to cache the calculating task；The JNI front-end proxy agent interface selects optimal clustered node, sends the calculating task；The optimal clustered node receives and executes the calculating task；The calculated result of the calculating task is obtained, and records properties data；Use the properties data of JSP simulation test webpage representation cluster.Performance test methods of the invention provide intuitive easy-to-use performance test interface by JSP page and JNI interface, for the acceleration computing cluster of non-Java；Using web browser as test interface, development difficulty is low, easy to use, can be realized cross-platform test.
Description
Technical field
The invention belongs to field of computer technology, in particular to a kind of clustering performance test method based on Java and it is
System.
Background technique
Http protocol be used to transmit information between Web browser and Website server, and http protocol is with clear-text way
Content is sent, the data encryption of any mode is not provided, if attacker has intercepted between Web browser and Website server
Transmitting message, so that it may information therein is directly understood, it is therefore, very dangerous using http protocol transmission privacy information.
In order to guarantee these private data energy encrypted transmissions, then Netscape devises Secure Socket Layer SSL (Secure
Sockets Layer) agreement, i.e. secure transport layer protocol TLS (Transport Layer Security), for being assisted to HTTP
The data of view transmission are encrypted, thus the HTTPS that has just been born.In simple terms, HTTPS agreement is by TLS/SSL+HTTP agreement
Building carries out encrypted transmission, the network protocol of authentication, than http protocol safety.
TLS/SSL agreement has the function of authentication, information encryption and completeness check, although making HTTPS opposite
HTTP has very big advantage, also just because of TLS/SSL agreement has used asymmetric key exchange technology in information ciphering process,
So that HTTPS performance and speed seriously reduce.
In traditional data calculation processing, it is hard that the computing capability of server depends on CPU, memory of server etc. itself
Part condition, and the computing capability of single server is limited after all；And some test assignments, such as RSA2048, ECC mainstream are non-
The encryption and decryption of symmetric encipherment algorithm calculates, shared key calculating scheduling algorithm task complexity is higher, larger to CPU consumption, not only
High load capacity occupies computing resource, also reduces the traffic handing capacity of software.
The bottleneck that break through, can use trunking mode: establish PC cluster method and system, undertake originally in separate unit
The magnanimity test assignment for needing to consume resource run on computer, to achieve the purpose that accelerate to calculate.Cluster is one group of phase
Computer that is mutually independent, being interconnected by high speed network, they constitute a group, and are managed with the mode of triangular web,
When one client and cluster interact, cluster seems an independent server.Every computer is considered as cluster in cluster
One node.
Above-mentioned computing cluster system is relatively high due to requiring in aspect of performance, and developer is frequently with C language or C++ language
Speech exploitation.Computing cluster system generally only provides the api interface of C language or C Plus Plus.In order to sufficiently show the workability of system
Can, need to do computing cluster performance test, but computing cluster system itself is not provided with and can intuitively display systems work
The visualization interface of performance, system performance testing are inconvenient.
Summary of the invention
In view of the above-mentioned problems, the present invention provides a kind of clustering performance test method based on Java, the performance test
Method includes:
It is instructed using JSP simulation test page creation performance test；
It is instructed according to the performance test, generates batch calculating task, JNI front-end proxy agent interface is called to cache the calculating
Task；
The JNI front-end proxy agent interface selects optimal clustered node, sends the calculating task；
The optimal clustered node receives and executes the calculating task；
The calculated result of the calculating task is obtained, and records properties data；
Use the properties data of JSP simulation test webpage representation cluster.
Further, the performance test includes encryption performance test, decryption performance test, signature performance test, sign test
Performance test and acquisition shared key performance test.
Further, the index of the performance test includes the calculating speed, system stability, load balancing energy of cluster
Power, wherein
The calculating speed of cluster is expressed as the task quantity of the execution of the cluster in the unit time；
Group system stability is expressed as the duration that cluster is capable of stable operation；
Cluster load balance ability is expressed as the poor ratio of most idle node and the resource utilization of most busy node in cluster.
Further, before the instruction using the page creation performance test of JSP simulation test, to JNI front-end proxy agent interface
It is initialized, the initialization includes specifying the algorithm of calculating task by setting parameter and notifying rear end cluster.
Further, described instructed using JSP simulation test page creation performance test includes the fortune that performance test is arranged
The row time.
Further, described to be instructed according to the performance test, generating batch calculating task includes being surveyed according to the performance
The timing of the runing time setting timer of examination, in the timing, circulation generates batch calculating task.
Further, the record properties data include that record clustering executes the task quantity completed and actual motion
Time.
Further, the record properties data include that the CPU usage of record clustering most idle node, memory make
With rate, and CPU usage, the memory usage of most busy node.
Further, the record properties data include record clustering operation duration.
The clustering performance test macro based on Java that the present invention also provides a kind of, which is characterized in that the performance test
System includes:
The JSP simulation test page shows the properties data of computing cluster for creating performance test instruction；
Java simulation test backstage generates batch calculating task, before calling JNI for instructing according to the performance test
Proxy interface is held to cache the calculating task；
JNI front-end proxy agent interface sends the calculating task, obtains described calculate and appoint for selecting optimal clustered node
The calculated result of business, and record the properties data；
Computing cluster receives and executes the calculating task.
Performance test methods of the invention: it by JSP page and JNI interface, is provided for the acceleration computing cluster of non-Java
Intuitive easy-to-use performance test interface；Using web browser as test interface, development difficulty is low, easy to use, Neng Goushi
Existing cross-platform test.
Other features and advantages of the present invention will be illustrated in the following description, also, partly becomes from specification
It obtains it is clear that understand through the implementation of the invention.The objectives and other advantages of the invention can be by specification, right
Pointed structure is achieved and obtained in claim and attached drawing.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 shows a kind of flow diagram of clustering performance test method based on Java according to the present invention；
Fig. 2 shows a kind of system structure diagrams of the clustering performance test macro based on Java according to the present invention；
Fig. 3 shows JSP simulation test page schematic diagram according to the present invention；
Fig. 4 shows the detailed process of the more idle node of selection according to the present invention；
Fig. 5 shows a kind of system for accelerating computing system based on the clustering of cipher card and CPU of the embodiment of the present invention
Frame diagram；
Fig. 6 is shown a kind of of the embodiment of the present invention and is accelerated to calculate system workflow based on the clustering of cipher card and CPU
Journey schematic diagram.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical solution in the embodiment of the present invention clearly and completely illustrated, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
The performance test to encrypting and decrypting, sign test of signing in asymmetric key exchange algorithm is not limited in the present invention, it can
Performance test suitable for the various algorithms run on PCI operation card, CPU, applied widely, the embodiment of the present invention is with right
The performance test of encrypting and decrypting, sign test of signing in asymmetric key exchange algorithm illustrates.
The clustering performance test method based on Java that the invention proposes a kind of, wherein clustering performance test index includes
It is the calculating speed of cluster, system stability, one or more in load balance ability.
The calculating speed of cluster is cluster main performance index, carrys out table with the task quantity that the cluster in the unit time executes
Show.By timer setting timing, circulation generates the batch calculating task of performance test, after calculating task is sent to
It holds computing cluster to execute, the operation result of calculating task is obtained, until timer timing arrives.The task that record clustering executes
Quantity and actual run time calculate the arithmetic speed of cluster.
Group system stability shows that testing tool sends calculating task to rear end computing cluster, and whether cluster can be steady
Fixed operation, is capable of the duration of stable operation with cluster to indicate.
Cluster load balance ability shows that testing tool high-volume issues calculating task, and each node calculating task of cluster is
No equilibrium is compared with the difference of most idle node in cluster and the resource utilization of most busy node to indicate, i.e. cluster load balance
Ability is equal to the difference of the resource utilization of most busy node and the resource utilization of most idle node again divided by most busy node
Resource utilization.Most busy node is resource utilization in clustered node (CPU, memory usage) maximum node, correspondingly,
Most idle node is the smallest node of resource utilization in clustered node.
A kind of clustering performance test method based on Java of the present invention, as shown in Figure 1, including the following steps.
Step 1: it is instructed using the page creation performance test of JSP simulation test.JSP, Java Server Pages, i.e.,
The java server page, a kind of dynamic web page technique standard are supported to be inserted into java applet section in traditional webpage html web page.
The performance test of rear end computing cluster includes encryption performance test, decryption performance test, signature performance test, sign test
Performance test and acquisition shared key performance test.
Encryption performance test refers to that test rear end computing cluster is encrypted using Encryption Algorithm (such as RSA, DSA algorithm)
Calculating speed, system stability and the load balance ability of operation.
The rivest, shamir, adelman and Diffie-Hellman of mainstream include RSA, DSA, DH, DHE, ECDH, ECDHE,
ECDSA algorithm.
Decryption performance test refers to test rear end computing cluster using Encryption Algorithm, be decrypted operation calculating speed,
System stability and load balance ability.
Signature performance test for testing rear end computing cluster using Encryption Algorithm, carry out signature operation calculating speed,
System stability and load balance ability.
Sign test performance test refers to that rear end computing cluster using Encryption Algorithm, carries out calculating speed, the system of sign test operation
Stability and load balance ability.
Shared key performance test is obtained for testing rear end computing cluster according to shared key algorithm, obtains shared key
Calculating speed, system stability and load balance ability.
The JSP simulation test page is established based on java server page JSP technology, creates performance on the simulation test page
Test instruction, every kind of performance test have corresponding icon (such as input frame, output box, to press on the JSP simulation test page
Button, drop-down menu), the result of performance test is also presented on the JSP simulation test page in the form of icon.
With java server page JSP develop WEB be it is cross-platform, can be in Linux, also can be in other operating systems
Upper operation.Java server page JSP is usually to execute in server end, and the result for returning to client is simple, only HTML
Text, client browser, that is, openable.
Before creation performance test instruction, first have to initialize JNI front-end proxy agent interface.Computing cluster system
Front-end proxy agent interface is using C/C++ language development, and the proxy interface is for C/C++ application call, java application
(such as Java simulation test backstage, java server page JSP) is not capable of calling the proxy interface, but the proxy interface is encapsulated
At JNI form, JNI front-end proxy agent interface is constituted, java application just can call directly the proxy interface.
Java local interface JNI (Java Native Interface) provides several API, may be implemented Java with
The communication (C and C++) of other language.
Illustratively, as shown in Fig. 2, rear end computing cluster is made of N number of node (computer), C/ is provided to external system
C++ front-end proxy agent interface, external system is by calling the C/C++ front-end proxy agent interface to obtain PC cluster service.The present invention makes
It is instructed with JSP simulation test page creation performance test, Java simulation test backstage is instructed according to test generates test batch
Calculating task, Java simulation test backstage can not call directly cluster C/C++ front-end proxy agent interface, in order to use PC cluster to take
The C/C++ front-end proxy agent interface encapsulation at JNI form, is constituted JNI front-end proxy agent interface by business, the present invention.
JNI front-end proxy agent interface initialization refers to that by initialization directive, assignment algorithm is arranged in the JSP simulation test page
Parameter notifies rear end computing cluster, the JSP simulation test page that certain algorithm to be called to calculate, and rear end computing cluster is according to instruction
Prepare corresponding queue, and generates corresponding public private key pair.Initialization algorithm does not generate task, and rear end computing cluster is initializing
When the public private key pair of response algorithm is generated using Openssl, cluster Openssl has corresponding function to generate key pair.?
On computer network, Openssl is the software library packet an of open source code, this packet can be used to be pacified in application program
Full communication avoids eavesdropping, while confirming the identity of other end connectionist.
Illustratively, as shown in figure 3, carrying out initiation parameter setting first on the JSP simulation test page, performance is selected
Then the initiation parameters such as arithmetic type, algorithm, traffic ID, curve mark, key digit used in testing send initialization
Instruct Java simulation test backstage；Java simulation test Background scheduling JNI front-end proxy agent interface carries out front-end proxy agent interface
Initialization, is arranged corresponding parameter.Wherein, arithmetic type includes cryptographic calculation, decryption operation, signature operation, sign test operation, obtains
Take shared key operation.
The parameter field of 1 initialization directive of table
| Algorithm field | Explanation | Corresponding field type |
| crypto_type | Algorithm | Int |
| nbit | Key digit | Int |
| openssl | Curve mark | Int |
| nworkid | Traffic ID | Int |
Above-mentioned initialization function includes parameter as shown in Table 1, and each parameter concrete meaning is as follows:
(1) algorithm: algorithm include mainstream rivest, shamir, adelman and Diffie-Hellman, such as RSA, DSA, DH,
DHE、ECDH、ECDHE、ECDSA。
(2) traffic ID: i.e. business is numbered, and rear end computing cluster can provide encryption and decryption operation for multiple front end service systems
Accelerate service, number is to distinguish different front-end businesses.
(3) curve identifies: ECDH, ECDHE, ECDSA algorithm are mixed with elliptic curve (ECC) algorithm, select these algorithms
When test, need reselection specific curve mark, comprising: secp192k1, secp224k1, secp224r1, secp384r1,
Secp521r1, brainpoolP256r1, brainpoolP384r1, brainpoolP512r1 etc..
(4) key digit: key digit is bigger, and the operations such as encryption and decryption are slower；When selecting RSA, DSA, DH, DHE scheduling algorithm,
Need to be arranged key digit, such as 512,1024,2048.
Illustratively, as shown in figure 3, Selecting operation type is cryptographic calculation, algorithms selection RSA, traffic ID, which is arranged, is
" 19 ", key digit select " 2048bit ", and " initialization " icon is clicked in curve mark selection " secp224k1 ", in " initialization
As a result " display " initializes successfully in output box！".
Each operation of performance test is all made of batch job mode, and performance test is arranged on the JSP simulation test page
Runing time (such as 100 seconds, 600 seconds or 1200 seconds), by operation the JSP simulation test page on respective icon, by performance
Test instruction is sent to Java simulation test backstage, and Java simulation test Background scheduling JNI front-end proxy agent interface, rear end calculates collection
Group executes batch calculating task.
Illustratively, as shown in figure 3, after above-mentioned example initializes successfully, in the JSP simulation test page " runing time
The performance test time is set in (second) " input frame, such as is set as " 600 " second, clicks " beginning " icon, and JSP simulation at this time is surveyed
It tries page creation and completes the test instruction of rsa encryption calculated performance, and send an instruction to Java simulation test backstage.
Step 2: instructing according to performance test, generates batch calculating task, and JNI front-end proxy agent interface caching is called to calculate
Task.
Java simulation test backstage receives the performance test instruction of the JSP simulation test page, according to above-mentioned performance test
Timer is arranged in time, and batch generates the calculating task of test, and every batch of calculating task is configurable, for example, according to every 10000
A calculating task is a batch, and recursive call JNI front-end proxy agent interface caches calculating task, until timer timing arrives.
Illustratively, initiation parameter is set when initialization: arithmetic type is cryptographic calculation, algorithms selection RSA, if
Setting traffic ID is " 19 ", and key digit selects " 2048bit ", curve mark selection " secp224k1 ", the JSP simulation test page
Setting the performance test time is " 600 " second, then Java simulation test backstage receives the secp224k1 curve mark of 2048bit
Rsa encryption cipher rounds execute instruction in 600 seconds, the timing that timer is arranged is 600 seconds, and every batch of calculating task is arranged
Number is 10000, starts timer, and Java simulation test backstage generates first 1000 calculating tasks, calls JNI front-end proxy agent
Interface sends first calculating task, and JNI front-end proxy agent interface caches first calculating task；Java simulation test backstage generates
After first calculating task, immediately begins to generate 1000 calculating tasks of second batch, JNI front-end proxy agent interface is called to send second
Calculating task is criticized, JNI front-end proxy agent interface is cached second batch calculating task, recycled with this, until 600 seconds timings of timer
It arrives.
Step 3: selecting optimal clustered node, sends calculating task.
JNI front-end proxy agent interface connects the clustered control service of rear end computing cluster, and clustered control service runs cluster
The clusters operation datas such as duration, cluster resource list return to JNI front-end proxy agent interface, and cluster resource list includes each node of cluster
Cpu resource utilization rate, the information such as memory source utilization rate.
JNI front-end proxy agent interface selects optimal clustered node according to cluster resource list, and the calculating task of caching is sent
To optimal clustered node.It is commented using cluster load balancing method according to the resource utilization of clustered node in cluster resource list
Idle clustered node ranking is selected, using the minimum node of resource utilization as optimal clustered node, distribution of computation tasks is given
Optimal clustered node realizes the dynamic load leveling between the clustered node of rear end.
Illustratively, compare two node resource utilization rates of A, B, select detailed process such as Fig. 4 institute of more idle node
Show, first determines whether node memory utilization rate is less than memory usage threshold value, for example, memory usage threshold value is 90%, when
When the memory usage of two nodes is below or is higher than, select the node that CPU usage is low in two nodes as free time degree
High node；When the memory usage of only one node meets threshold value, judge whether the CPU usage of the node meets
90%: if satisfied, the then node that the node is high as free time degree；If not satisfied, then selecting CPU usage in two nodes low
The node node high as free time degree, if the CPU usage of two nodes is equal, further the memory of relatively two nodes makes
With rate, the lower node of memory usage is selected as optimal clustered node.
Step 4: cluster receives and executes calculating task.
Optimal clustered node receives calculating task, caching calculating task to task buffer queue, and execution unit delays from task
It deposits queue and reads calculating task progress operation, and calculated result is returned into JNI front-end proxy agent interface.
Cipher card is a kind of PCI operation card, is high performance fundamental cryptographic equipment, is the important component of cryptographic hardware platform
One of, be widely used in a variety of encryption devices such as signature verification service device, vpn gateway, it is main complete encryption, decryption, signature,
The functions such as sign test.
Illustratively, there are two types of the calculating devices of clustered node execution calculating task, it is CPU and cipher card respectively, corresponds to
Worker thread be CPU worker thread, cipher card worker thread respectively.CPU worker thread, cipher card active line is respectively started
Journey, worker thread distribute calculating task according to certain calculating task dispatching method, to CPU or cipher card, CPU worker thread,
Cipher card worker thread executes calculating task respectively, result cache queue is written in task computation result, and return to calculated result
Give JNI front-end proxy agent interface.
Step 5: obtaining the calculated result of calculating task, and records properties data.
Java simulation test Background scheduling JNI front-end proxy agent interface obtains the calculated result of calculating task, and record clustering
The task quantity completed and actual run time are executed, the arithmetic speed of this performance test cluster is calculated.
Illustratively, the JSP simulation test page setting performance test time is 600 seconds, issues performance test instruction, Java
Simulation test backstage timing be 600 seconds, circulation production batch calculating task, rear end computing cluster in timing always
It continuously carries out calculating task, returns to task computation result to JNI front-end proxy agent interface, before Java simulation test Background scheduling JNI
It holds proxy interface to obtain the calculated result of calculating task, and records in real time and update cluster and execute the task quantity and reality completed
Runing time, until timing arrives.Then, the cluster of Java simulation test backstage final entry executes completion to timing
Task quantity is 166800, and actual test runing time is timer timing, i.e., 600 seconds, calculates this performance test collection
The arithmetic speed of group is that the task quantity that cluster executes completion is 16 680000 divided by actual test runing time 600 seconds, is equal to
27 800 (secondary/seconds).
Java simulation test Background scheduling JNI front-end proxy agent interface obtains rear end computing cluster resource, and using list, (CPU makes
With rate, memory usage), and calculate current cluster load balance ability.Cluster load balance ability includes cpu load equilibrium degree
With memory load balancing degrees.
Illustratively, the JSP simulation test page setting performance test time is 600 seconds, issues performance test instruction, Java
Simulation test backstage timing be 600 seconds, circulation production batch calculating task, rear end computing cluster in timing always
Continuously carry out calculating task, Java simulation test Background scheduling JNI front-end proxy agent interface timing acquisition rear end computing cluster resource
Using list, and update the resource utilization of most idle node and most busy node.In renewal process, once discovery exists than working as
The node updates are then most idle node by the also low node of the resource utilization of preceding most idle node, similarly, once discovery is deposited
It is then most busy node by the node updates in the node also higher than the resource utilization of currently most busy node.Such as when
When Java simulation test background timer timing is 10 seconds, the CPU usage of most idle node is 67%, and memory usage is
61%, the CPU usage of most busy node is 75%, memory usage 56%, when Java simulation test background timer meter
When be 30 seconds when, discovery one node cpu utilization rate be 64%, memory usage 54%, then using the node as most idle section
Point, when Java simulation test background timer timing is 90 seconds, one node cpu utilization rate of discovery is 80%, memory usage
It is 60%, then using the node as most busy node, until AVA simulation test background timer timing arrives, most idle section
Point does not update again with most busy node.So, then, the CPU usage of most idle node is 64% to timer timing,
Memory usage is 54%, and the CPU usage of most busy node is 80%, memory usage 60%, and it is balanced to calculate cpu load
Difference of the degree equal to 80% and 64% calculates difference of the memory load balancing degrees equal to 60% and 54% and is in divided by 80%, i.e., 20%
60%, i.e., 10%.
Java simulation test Background scheduling JNI front-end proxy agent interface obtains rear end computing cluster operation duration, obtains cluster
This test performance index of system stability.
Illustratively, the task that executes in batches is generated by JSP page triggering, records the time started, persistent loop, which generates, appoints
Business, setting execute time long enough or are not provided with the time, task execution effect are monitored, when current task cannot be executed or be held out
When the case where row effect is substantially reduced, the record end time.By end time and time started, Mass production task
Density (treating capacity of the single clustered node in the unit time), assesses the stability of group system.Further, monitoring task is held
Market condition can be obtained by executing task feedback result in the dynamic change of front end JSP page, can also be by checking collection
Group node execution journal searches and executes stopping or the time-consuming timing node obviously increased.It is possible to further repeatedly survey in batches
Stability is tried, different task density is set every time, accurately to test out system for the stability of different task density.
Step 6: the properties data of JSP simulation test webpage representation cluster are used.
The JSP simulation test page not only creates performance test instruction, can also intuitively show the performance data of cluster, to survey
The friendly test interface of examination personnel offer and simply and easily test method.
Illustratively, Fig. 3 shows effect of the performance data in JSP simulation test page presentation of cluster, as seen from the figure,
Performance exports a column and shows three cluster arithmetic speed, cluster operation duration, cluster load balance ability performance indicators, that is, collects
Group operatione speed is 27 800 times/second, cluster operation duration be 5 hours 20 days 20 points 18 seconds, memory load balancing degrees are 10%,
Cpu load equilibrium degree is 20%.
A kind of clustering performance test method based on Java of the present invention devises a kind of performance test methods of computing cluster,
Visual performance test methods are provided using the JSP simulation test page, intuitively shows cluster properties data, facilitates test
Personnel use and operate, and improve the efficiency of performance test.
The present invention also provides a kind of clustering performance surveys based on Java being adapted with above-mentioned clustering performance test method
Test system, as shown in figure 3, the system include the JSP simulation test page, Java simulation test backstage, JNI front-end proxy agent interface and
Rear end computing cluster.
The JSP simulation test page is used for JNI front-end proxy agent interface initialization, and creation includes encryption performance test, decryption
Performance test, the signature performance tests such as performance test and sign test performance test instruction, sending performance test instruction are simulated to Java
Test background shows the properties data of computing cluster.
Java simulation test backstage provides the background process for being directed to the JSP simulation test page, receives the JSP simulation test page
Instruction, instructs according to performance test, generates batch calculating task, and JNI front-end proxy agent interface is called to cache calculating task, obtains meter
The calculated result of calculation task, and record properties data.
JNI front-end proxy agent interface receives the calculating knot of calculating task for selecting optimal clustered node to send calculating task
Fruit.
Rear end computing cluster is for receiving and executing calculating task.Rear end computing cluster includes calculating service and clustered control
Subsystem, wherein calculate service subsystem and be made of computing module, high-speed communication module, data statistics module；Clustered control clothes
Business manages all clustered nodes, provides cluster resource list, cluster resource list includes the CPU usage and memory of clustered node
Utilization rate.
Computing module is responsible for executing the calculating task of various algorithms；High-speed communication module mainly realizes mission bit stream and calculating
As a result quick transmission, is responsible for the calculating task of receiving front-end, and returns to calculated result；Data statistics module is responsible for real-time statistics
Task data including unfinished calculating task number, pre-estimation evaluation time, and is sent to cluster by the task data of present node
Control service.
A kind of above-mentioned clustering performance test macro based on Java, devises the performance testing tool of a set of computing cluster,
The visual performance test page is provided using the JSP simulation test page, intuitively shows cluster properties data, facilitates test
Personnel use and operate, and improve the efficiency of performance test；It is tested for the property using Java and supports cross-platform test.
Illustratively computing cluster is accelerated to be briefly described the encryption of the embodiment of the present invention below.The cluster is one
Kind accelerates computing system based on the clustering of cipher card and CPU, and system architecture includes hardware layer, hardware interface encapsulated layer, rear end
The levels such as computing cluster layer, front-end proxy agent interface layer, calling service layer, as shown in Figure 5.
Hardware layer, which provides to calculate, services supported hardware, including the hardware devices such as PCI operation card and CPU.Cipher card is
A kind of PCI operation card is to emphasize high performance fundamental cryptographic equipment, is one of the important component of cryptographic hardware platform, answers extensively
In a variety of encryption devices such as signature verification service device, vpn gateway.
Openssl of the hardware interface encapsulated layer to cipher card based on Openssl calculating interface and based on Intel optimization
Interface carry out secondary encapsulation, provide support cipher card and CPU collaboration operation external call interface, for front-end proxy agent interface layer into
Row, which calculates, to be called.On the computer network, Openssl is the software library packet an of open source code, and application program can be used
This packet avoids eavesdropping, while confirming the identity of other end connectionist to securely communicate.It is based on Intel optimization
Refer to the optimization for making full use of the new commands such as AVX2 instruction set and MULX, ADCX, ADOX, RORX, RDSEED to carry out.
Rear end computing cluster layer is responsible for the transmitting-receiving management of calculating task, hardware enclosure interface is called to execute calculating task, and
Calculated result is cached, front-end proxy agent interface is sent to.Calculating task includes the tasks such as encrypting and decrypting, signature sign test.
Rear end computing cluster layer is mainly made of calculating service, clustered control service, information collection service, calculates services package
Include Thrift communication service frame, task transceiver module, computing module.Thrift can be used as high performance communication middleware,
Support data (object) serializing and a plurality of types of RPC services.RPC (Remote Procedure Call Protocol) clothes
Business refers to (RFC-1831) remote procedure call protocol, and one kind requesting service by network from remote computer program, without
It should be understood that the agreement of underlying network technology.
Task transceiver module is responsible for receiving front-end and acts on behalf of all kinds of calculating tasks sent by exchange network, is responsible for calculate
As a result front-end proxy agent is returned to by exchange network, the interface function that the API communication initialization and API for responding front-end proxy agent are closed.
Caching mechanism is used to received calculating task and calculated result, including without lock task queue and without lock result queue.
Front-end proxy agent interface layer is responsible for the inquiry of clustered node, the caching of calculating task and transmission, the reception of calculated result,
Caching and inquiry, the tasks such as control of task traffic volume.
Calling service layer is responsible for calling front-end proxy agent interface, sends calculating task, fetches task action result；It further include hair
Identity after identity registration and execution before sending calculating task is nullified.
The present invention is a kind of to accelerate computing system to be broadly divided into front end services and rear end collection based on the clustering of cipher card and CPU
Group calculates two parts.
Front end services part is made of front-end proxy agent interface layer, calling service layer, is deployed on front-end server, mainly with
Front-end business is docked, and calculating task is issued rear end computing cluster by interface, and return to calculated result.
Front end services part includes front-end proxy agent software and operation system, because common server system has had business system
System, so operation system is not emphasis of the invention.
PC cluster part is made of hardware layer, hardware interface encapsulated layer, rear end computing cluster layer, is deployed in rear end cluster
On server, include six modules such as high-speed communication, task transmitting-receiving, cipher card operation, CPU operation, log recording.
PC cluster part includes hardware and software, and the difference of hardware and common server is, in existing server hardware
On the basis of cipher card is installed.
Front end services part is docked with PC cluster part, and calculating task is issued rear end PC cluster by interface
Part, and return to calculated result.
PC cluster part includes high-speed communication, task transmitting-receiving, information collection, password card interface, CPU operation interface, meter
Calculate the modules such as task schedule, log recording.
Illustratively, a kind of that computing system is accelerated based on the clustering of cipher card and CPU, as shown in fig. 6, the accelerometer
The front end of calculation system is equipped with M front-end server, the M front-end server that be the 1st, 2,3 respectively ..., and each front-end server includes
Operation system and front-end proxy agent.Rear end is equipped with computing cluster, database and cluster management subsystem.Computing cluster is by N number of calculating
Node composition, N number of calculate node server that is the 1st, 2,3 respectively ..., each calculate node server is mutually indepedent, passes through high speed
The network interconnection, they constitute a group, and are managed with the mode of triangular web.Each calculate node includes calculating clothes
Business, information collection service and clustered control service.Database and cluster management subsystem can be deployed in independently of clustered node
Individual server on, can also be deployed on clustered node server, pass through software interface and clustered node and carry out data friendship
Mutually.Accelerate the workflow of computing system as follows:
(1) front end service system calls front-end proxy agent interface to issue computation requests.
(2) front-end proxy agent receives and caches calculating task, the clustered node clustered control service of connection configuration setting.
(3) clustered control services return node the Resources list.
(4) front-end proxy agent selects optimal rear end calculate node server and connects.
(5) rear end calculate node caches calculating task.
(6) rear end calculate node returns to front-end proxy agent task and receives successfully.
(7) front-end proxy agent interface returns to front end service system calculating task call result.
(8) rear end calculate node executes calculating task.
(9) rear end calculate node returns to front-end proxy agent task computation result.
(10) front-end proxy agent caches task computation result.
(11) front end service system calls front-end proxy agent interface to carry out task computation result queries.
(12) front-end proxy agent interface returns to front end service system task computation result.
(13) information collection service timing acquiring nodal information (resource utilization, task situation etc.).
(14) information collection service connection clustered control service sends collected nodal information.
(15) all nodal informations of cluster are read in information storage service connection clustered control service, and number is written in cluster information
According to library.
(16) it cluster management subsystem reading database data and shows.
Although the present invention is described in detail referring to the foregoing embodiments, those skilled in the art should manage
Solution: it is still possible to modify the technical solutions described in the foregoing embodiments, or to part of technical characteristic into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The spirit and scope of scheme.
Claims (10)
1. a kind of clustering performance test method based on Java, which is characterized in that the performance test methods include:
It is instructed using JSP simulation test page creation performance test；
It is instructed according to the performance test, generates batch calculating task, described calculate of JNI front-end proxy agent interface caching is called to appoint
Business；
The JNI front-end proxy agent interface selects optimal clustered node, sends the calculating task；
The optimal clustered node receives and executes the calculating task；
The calculated result of the calculating task is obtained, and records properties data；
Use the properties data of JSP simulation test webpage representation cluster.
2. performance test methods according to claim 1, which is characterized in that the performance test includes that encryption performance is surveyed
Examination, decryption performance test, signature performance test, sign test performance test and acquisition shared key performance test.
3. performance test methods according to claim 1, which is characterized in that the index of the performance test includes cluster
It is calculating speed, system stability, one or more in load balance ability, wherein
The calculating speed of cluster is expressed as the task quantity of the execution of the cluster in the unit time；
Group system stability is expressed as the duration that cluster is capable of stable operation；
Cluster load balance ability is expressed as the poor ratio of most idle node and the resource utilization of most busy node in cluster.
4. performance test methods according to claim 1, which is characterized in that described to use JSP simulation test page creation
Before performance test instruction, JNI front-end proxy agent interface is initialized, the initialization includes calculating by the way that setting parameter is specified
The algorithm of task simultaneously notifies rear end cluster.
5. performance test methods according to claim 1-4, which is characterized in that described to use JSP simulation test
Page creation performance test instruction includes: that the runing time of performance test is arranged.
6. performance test methods according to claim 5, which is characterized in that it is described to be instructed according to the performance test, it is raw
Calculating task includes: the timing that timer is arranged according to the runing time of the performance test in batch, in the timing
In time, circulation generates batch calculating task.
7. performance test methods according to claim 1, which is characterized in that the record properties data include record
Cluster executes the task quantity completed and actual run time.
8. performance test methods according to claim 1 or claim 7, which is characterized in that the record properties data include:
CPU usage, the memory usage of record clustering most idle node, and CPU usage, the memory of most busy node use
Rate.
9. performance test methods according to claim 1 or claim 7, which is characterized in that the record properties data include
Record clustering operation duration.
10. a kind of clustering performance test macro based on Java, which is characterized in that the Performance Test System includes:
The JSP simulation test page shows the properties data of computing cluster for creating performance test instruction；
Java simulation test backstage generates batch calculating task, calls the front end JNI generation for being instructed according to the performance test
It manages interface and caches the calculating task；
JNI front-end proxy agent interface sends the calculating task, obtains the calculating task for selecting optimal clustered node
Calculated result, and record the properties data；
Computing cluster receives and executes the calculating task.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910385999.9A CN110188022A (en) | 2019-05-09 | 2019-05-09 | A kind of clustering performance test method and system based on Java |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910385999.9A CN110188022A (en) | 2019-05-09 | 2019-05-09 | A kind of clustering performance test method and system based on Java |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110188022A true CN110188022A (en) | 2019-08-30 |
Family
ID=67715939
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910385999.9A Pending CN110188022A (en) | 2019-05-09 | 2019-05-09 | A kind of clustering performance test method and system based on Java |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110188022A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110633579A (en) * | 2019-09-02 | 2019-12-31 | 上海同在互联网科技有限公司 | Bidding file online decryption result prediction method, system, device and storage medium |
| CN112346947A (en) * | 2020-11-16 | 2021-02-09 | 百度在线网络技术（北京）有限公司 | Performance detection method and device, electronic equipment and computer readable medium |
| CN113434374A (en) * | 2021-06-17 | 2021-09-24 | 华东师范大学 | Multithreading POSIX standard-based real-time operating system memory management algorithm performance test method and system |
| US11954013B2 (en) | 2020-11-16 | 2024-04-09 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method of testing applet performance, electronic device, and computer-readable medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100333072A1 (en) * | 2009-06-30 | 2010-12-30 | Sap Ag | Integrated performance and load testing tool for application servers |
| CN103581247A (en) * | 2012-07-30 | 2014-02-12 | 杭州洱海科技有限公司 | Method for testing distributed type Web based on cloud computing environment |
| CN104461856A (en) * | 2013-09-22 | 2015-03-25 | 阿里巴巴集团控股有限公司 | Performance test method, device and system based on cloud computing platform |
| CN104520814A (en) * | 2012-08-07 | 2015-04-15 | 超威半导体公司 | System and method for configuring cloud computing systems |
| CN106850321A (en) * | 2017-04-05 | 2017-06-13 | 无锡华云数据技术服务有限公司 | A kind of simulated testing system of cluster server |
-
2019
- 2019-05-09 CN CN201910385999.9A patent/CN110188022A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100333072A1 (en) * | 2009-06-30 | 2010-12-30 | Sap Ag | Integrated performance and load testing tool for application servers |
| CN103581247A (en) * | 2012-07-30 | 2014-02-12 | 杭州洱海科技有限公司 | Method for testing distributed type Web based on cloud computing environment |
| CN104520814A (en) * | 2012-08-07 | 2015-04-15 | 超威半导体公司 | System and method for configuring cloud computing systems |
| CN104461856A (en) * | 2013-09-22 | 2015-03-25 | 阿里巴巴集团控股有限公司 | Performance test method, device and system based on cloud computing platform |
| CN106850321A (en) * | 2017-04-05 | 2017-06-13 | 无锡华云数据技术服务有限公司 | A kind of simulated testing system of cluster server |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110633579A (en) * | 2019-09-02 | 2019-12-31 | 上海同在互联网科技有限公司 | Bidding file online decryption result prediction method, system, device and storage medium |
| CN112346947A (en) * | 2020-11-16 | 2021-02-09 | 百度在线网络技术（北京）有限公司 | Performance detection method and device, electronic equipment and computer readable medium |
| WO2022100075A1 (en) * | 2020-11-16 | 2022-05-19 | 百度在线网络技术（北京）有限公司 | Method and apparatus for performance test, electronic device and computer-readable medium |
| CN112346947B (en) * | 2020-11-16 | 2023-08-01 | 百度在线网络技术（北京）有限公司 | Performance detection method, device, electronic equipment and computer readable medium |
| US11954013B2 (en) | 2020-11-16 | 2024-04-09 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method of testing applet performance, electronic device, and computer-readable medium |
| CN113434374A (en) * | 2021-06-17 | 2021-09-24 | 华东师范大学 | Multithreading POSIX standard-based real-time operating system memory management algorithm performance test method and system |
| CN113434374B (en) * | 2021-06-17 | 2022-10-25 | 华东师范大学 | Method and system for testing performance of memory management algorithm of real-time operating system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aujla et al. | BloCkEd: Blockchain-based secure data processing framework in edge envisioned V2X environment | |
| CN110188022A (en) | A kind of clustering performance test method and system based on Java | |
| Yala et al. | Latency and availability driven VNF placement in a MEC-NFV environment | |
| Elgendy et al. | An efficient and secured framework for mobile cloud computing | |
| Luo et al. | Cloud-based information infrastructure for next-generation power grid: Conception, architecture, and applications | |
| Schafer et al. | Tasklets:" better than best-effort" computing | |
| US20160300142A1 (en) | System and method for analytics-driven sla management and insight generation in clouds | |
| Alam et al. | Blockchain, fog and IoT integrated framework: review, architecture and evaluation | |
| CN106254121A (en) | A kind of automatization disposes and the method for management large data sets group | |
| CN109857542A (en) | Calculate power resource adjustments method, system and device | |
| CN110213338A (en) | A kind of clustering acceleration calculating method and system based on cryptographic calculation | |
| Li et al. | GSPN-based reliability-aware performance evaluation of IoT services | |
| CN116305136A (en) | Source audit trail for micro-service architecture | |
| Flores et al. | Mobile cloud messaging supported by xmpp primitives | |
| Ouedraogo et al. | A cost-effective approach for end-to-end QoS management in NFV-enabled IoT platforms | |
| Yamanaka et al. | Proximity-aware IaaS in an edge computing environment with user dynamics | |
| Venâncio et al. | Beyond VNFM: Filling the gaps of the ETSI VNF manager to fully support VNF life cycle operations | |
| Yang et al. | Delay-aware secure computation offloading mechanism in a fog-cloud framework | |
| CN110213043A (en) | A kind of clustering functionality test method based on Java | |
| CN112631800A (en) | Kafka-oriented data transmission method and system, computer equipment and storage medium | |
| Saab et al. | Energy efficiency in mobile cloud computing: Total offloading selectively works. does selective offloading totally work? | |
| CN114785526B (en) | Multi-user multi-batch weight distribution calculation and storage processing system based on block chain | |
| Nawrocki et al. | Adaptive context‐aware service optimization in mobile cloud computing accounting for security aspects | |
| Zhao et al. | Minimize the make-span of batched requests for FPGA pooling in cloud computing | |
| Hu et al. | Joint Content and Radio Access for the Internet of Things: A Smart-Contract-Based Trusted Framework |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190830 |