CN110737564A

CN110737564A - VmWare-based virtual machine performance monitoring method and system

Info

Publication number: CN110737564A
Application number: CN201810798743.6A
Authority: CN
Inventors: 朱国骏; 符健; 李铂
Original assignee: Individual
Current assignee: Individual
Priority date: 2018-07-19
Filing date: 2018-07-19
Publication date: 2020-01-31

Abstract

The invention discloses VmWare-based virtual machine performance monitoring methods and systems, which comprise the following steps that a virtual machine Server end starts Socket monitoring in a Server thread pool, a link request of a Client end is monitored, a test machine Client end uses a TCP, UDP or TCP/UDP mixed network protocol to perform Socket communication with a virtual machine Server end, a test task is executed, a test result is collected and sent to a main control machine, the main control machine Manager sends a test command to a Client end test machine, distributed test results are collected and counted, and the main control machine Manager sends an RPC request to a virtual machine, obtains various performance indexes of the virtual machine and displays the performance indexes in a chart form.

Description

VmWare-based virtual machine performance monitoring method and system

Technical Field

The invention relates to the field of virtual machine performance monitoring, in particular to a Vmware virtual machine performance monitoring method and system.

Background

In recent years, powerful and relatively cheap multiprocessor servers have become the mainstream of computer hardware development, but in many cases, the performance of the hardware cannot be fully utilized, and in recent years, the insufficiency is further aggravated with the popularization of multi-core processors, which makes the industry more and more attach importance to virtualization technology.

Currently, SPEC used in host benchmarking is only focused on checking a single workload, but cannot obtain the performance of a server supporting multiple virtual machine systems or simultaneously supporting multiple workloads on servers, more appropriate virtual machine benchmarking should check the performance of the server using real and diverse workloads running on multiple operating systems, and the benchmarking should generate easily understood indexes.

The reason for these differences is determined by the technical features of the virtualization itself:

in the field of computer science, virtualization represents an abstraction of computing resources and is not limited to the concept of virtual machines, in general virtualization technologies fall into several broad categories:

1) platform Virtualization (Platform Virtualization), Virtualization for computers and operating systems;

2) resource Virtualization (Resource Virtualization), Virtualization for specific system resources, such as memory, storage, network resources, etc.

3) Application Virtualization (Application Virtualization) includes emulation, simulation, interpretation techniques, and the like.

Secondly, platform virtualization technology can be subdivided into full virtualization, partial virtualization, hardware-assisted virtualization, etc. [3 ].

The interaction between the operating system and the real hardware can be seen as being performed through predefined hardware interfaces, and the fully virtualized VMM provides all interfaces in a manner of completely simulating the hardware, which typically represents Vmware, VirtualBox and the like.

Partial virtualization only simulates part of bottom-layer hardware and only provides simulation of critical computing components or instruction sets, so that a guest operating system cannot run in a virtual machine without modification, other programs may also need to be modified, and a popular local virtualization VMM has Xen, Hyper-V.

Therefore, the whole performance of the virtual machine can be affected differently by full virtualization and partial virtualization, and Duchun Yang et al compare the performance of several different virtualization technologies in a high-performance computing system and a clustering technology through a NPB symmetric multiprocessing performance test, test the serial and parallel performance on a single node based on the NPB, and compare the serial and parallel performance with the original NPB standard to draw the conclusion that no virtualization technologies can perfectly achieve the performance of a native system and roughly analyze the performance of different VMMs in different standards.

Therefore, the method for testing the virtual machine, which is suitable for different specific enterprise applications, is very significant in research and practical value, and the technology can be widely applied to management of the virtual machine of the computing resource of the cloud computing platform and evaluation of the number and type of deployed applications on the virtual machine in .

The VMM is an abstraction and simulation of the real computing environment, the VMM needs to allocate sets of data structures to each virtual machine to manage their state, including the full set of registers of the virtual processor, the usage of physical memory, the state of the virtual devices, etc. when the VMM schedules a virtual machine, it restores its partial state to the host system.

Enterprises or scientific research institutions usually pay more attention to the performance problem of virtualization technologies when facing a wide variety of virtualization technologies with various characteristics. Firstly, whether the performances provided by various virtualization technology implementations are greatly different or not; second, there is a difference in performance between applications running on a virtual machine and running directly on a physical machine.

Lorenzo Martigoni et al compare differences between different virtual machines and physical machines by performing a special protocol-based fuzzy test and a difference analysis in user-mode and kernel-mode operating modes in physical machine and virtual machine environments, respectively, but this test method places more of the center of gravity on the differences in the functions of the physical machine and the virtual machine, and the differences in performance are less involved.

Because the virtual machine system has poor performance in the aspect of virtualized disk storage due to the use of a traditional storage mode, Dongsheng Yin et al provide I/O storage optimization schemes based on SSD and a shared RAID array by comparing virtual and real storage experiments of a VMware vSphere4.0 virtual machine.

The juanhua Che and the like perform black box test on VMM tools of three open sources of OpenVZ, Xen and KVM, which respectively adopt virtualization Based on Container (Container-Based), local virtualization and full virtualization strategies by using standard mechanism test programs of CPU2006, RAMSPEED, Bonnie + + and the like, wherein test indexes of the black box test include CPU, memory, disk IO, network of a virtual machine and system operation context switching operation frequency of a more detailed micro-operation level, and finally, the excellent abuse of code design is analyzed from the perspective of white box test through test result data. The research angle is more that the test analysis is performed from the viewpoint of a VMM designer, better suggestions are provided for improving the design of the VMM, and the guiding significance is limited under the practical application environment.

Almari et al compared the performance class extensibility of Orcale databases in both Orcale VM and Vmware ESxi virtualization environments to verify whether the statement that Orcale VM declared to be several times the Vmware ESxi performance in the market is correct, and the results of experiments show that the selection of virtualized products should be based on specific application requirements, especially the types of ESxi that the application of databases should be based on read-write bandwidth and transaction rate.

The virtualization technology is used as an important component of cloud computing, and Dioniio Leite et al researches performance difference of the virtualization technology in a cloud environment, performs load test on a host machine and a virtual machine under the condition of the same load, and quantifies the maximum number of Xen and KVM in parallel operation under the condition of .

Therefore, the research on virtualization performance at home and abroad can be seen to focus on the evaluation and comparison of the performance of different VMMs, the comparison between virtual machines and physical machines is limited to certain specific applications, so the problem of performance comparison between virtual machines and physical machines is to be researched .

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provides VmWare-based virtual machine performance monitoring methods and systems.

Another objective of the invention is to provide a monitoring system for virtual machine performance.

The purpose of the invention is realized by the following technical scheme:

VmWare-based virtual machine performance monitoring method, comprising the following steps in sequence:

s1, a virtual machine Server end starts socket monitoring in a Server thread pool, and monitors a link request of a client;

s2, the Client end of the tester uses a TCP (transmission control protocol) network protocol, a UDP (user datagram protocol) network protocol or a TCP/UDP (transmission control protocol/user datagram protocol) hybrid network protocol to perform Socket communication with the server end of the virtual machine, execute a test task, collect a test result and send the test result to the main control computer;

s3, the main control computer Manager sends a test command to the client test machine, and collects and counts distributed test results;

and S4, the main control machine Manager sends an RPC request to the virtual machine to obtain various performance indexes of the virtual machine, and the performance indexes are displayed in a chart form.

In step S3, the collecting and counting distributed test results specifically includes that the main control computer Manager collects test result data of the client test machine times every 3S, averages different collected data, and counts the current Socket transmission rate.

In step S3, the collecting distributed test results specifically includes the following steps:

a. the RPC client constructs a udp request through a Socket according to the IP address of the target server and the default program port 111 of the portmapper, and sends the request to a portmapper application program of the target machine to inquire the program port number of the rstatd;

b. after receiving the request, the portmapper application queries a mapping table between the port and the application, and returns the port number of the rstatd monitoring program to the RPC client;

c. after the RPC client obtains the port number of the rstatd program, a Socket reconstructs a new Udp request according to the original target IP address and the rstatd at the port number, and sends a request service to the rstatd monitoring program;

d. after the Rstatd monitor receives the service request, the performance data collected by the current performance counter is returned.

In step S4, the obtaining of each performance index of the virtual machine and the displaying in the form of a chart specifically include:

a. storing the performance data in the test process in a DataTable object, and updating the data transmission acquired by Linux remote performance monitoring in real time;

b. after the test is finished, the performance data in the DataTable is counted and summarized to obtain final performance test data;

c. then, defining the style of data presentation by using XSL, and writing the test data into an XML file according to rules determined by ;

d. finally, the XML formatted data file is converted into a user-friendly Html page by the xsl compiledtransform class under the system.

The other purpose of the invention is realized by the following technical scheme:

virtual machine performance monitoring system comprises a virtual machine Server end, a test machine Client end and a main control machine Manager which are connected with each other in pairs, wherein the virtual machine Server end starts Socket monitoring in a Server thread pool and monitors a link request of a Client, the test machine Client end performs Socket communication with the virtual machine Server end by using network protocols such as TCP, UDP and the like, executes a test task, collects a test result and sends the test result to the main control machine, the main control machine Manager sends a test command to a Client test machine, collects and counts the test result, sends an RPC request to the virtual machine, acquires various performance indexes of the virtual machine and displays the performance indexes in a chart form.

Compared with the prior art, the invention has the following advantages and beneficial effects:

1. the virtualization technology studied in the present invention refers to performance evaluation under platform virtualization technology, and provides users with an abstract, systematic , simulated computing environment (referred to as Virtual Machine) by using a Control Program (also referred to as Virtual Machine Monitor or Hypervisor), an operating system running in a Virtual Machine is referred to as Guest operating system (Guest OS), and an operating system running a Virtual Machine Monitor is referred to as Host operating system (Host OS).

2. The invention focuses on the development and implementation of network performance evaluation under a virtualization environment, and designs a network performance testing framework which is universal under the environment of a physical machine and a virtual machine. The software has the following characteristics: the system supports multi-network transmission protocols, multi-thread distributed test and remote real-time monitoring of system performance indexes. The method specifically comprises the following steps:

(1) statistics of VMM-based network transmission rates

Different network transmission protocols used by different applications are different, so that a network performance detection mechanism for virtual machines under different transmission protocols is provided, and three network transmission modes of TCP, UDP and TCP/UDP mixing are supported at present.

(2) Multi-threaded distributed testing

The test tasks are deployed into different distributed systems and managed by a Host system , each test tasks support multi-thread parallel operation, the network transmission performance of the virtual machine is counted in real time, and the test tasks are displayed in a chart mode.

(3) System remote performance monitoring

In the process of executing the test task, the main control computer can perform real-time system monitoring on each performance index of the tested machine so as to assist in evaluating the network performance of the virtual machine.

Drawings

FIG. 1 is a flow chart of remote system call of RPC client to obtain Linux performance parameters;

FIG. 2 is a flow chart of a report data presentation process;

FIG. 3 is a schematic diagram of collecting distributed test data;

fig. 4 is a schematic structural diagram of an virtual machine performance monitoring system according to the present invention.

Detailed Description

The present invention will be described in further detail with reference to examples and drawings, but the embodiments of the present invention are not limited thereto.

s3, the main control machine Manager sends a test command to the client test machine, and collects and counts distributed test results, wherein the collecting and counting of the distributed test results are specifically that the main control machine Manager collects times of test result data of the client test machine every 3s, averages different collected data, and counts the current Socket transmission rate;

as shown in fig. 3, the gathering of the distributed test results specifically includes the following steps:

d. after the Rstatd monitoring program receives the 'request for service', the performance data collected by the current performance counter is returned;

in fig. 3, (1), (2), (3), (4) correspond to the above substeps a, b, c, d, respectively;

s4, the main control machine Manager sends an RPC request to the virtual machine to obtain various performance indexes of the virtual machine, and the performance indexes are displayed in a chart form;

as shown in fig. 2, the obtaining of various performance indexes of the virtual machine and the displaying in the form of a chart specifically include:

As shown in FIG. 4, the virtual machine performance monitoring system includes a virtual machine Server end, a test machine Client end, and a host computer Manager connected with each other two by two, where the virtual machine Server end starts Socket monitoring in a Server thread pool and monitors a link request of a Client, the test machine Client end performs Socket communication with the virtual machine Server end using TCP, UDP, and other network protocols, executes a test task, collects a test result, and sends the test result to the host computer, the host computer Manager sends a test command to the Client test machine, collects and counts the test result, and sends an RPC request to the virtual machine, obtains various performance indexes of the virtual machine, and displays the performance indexes in the form of a graph.

The monitoring system manages the distributed testing machine through the main control machine and distributes testing tasks to the client testing machine; the client establishes connection with the virtual machine through the local area network, executes a test task, and returns a real-time test result to the main control machine. And the main control machine sends an RPC request to the vmm virtual machine to obtain various performance indexes of the virtual machine. The various components of the system are described in detail below with reference to FIG. 4:

virtual machine Server terminal

Starting an operating system rstat port monitor ( -like Linux system is installed by default, if an rstat port monitor tool is not installed, firstly, downloading a Tar compressed file of the rstat from an official website, after decompressing the Tar file, compiling, configuring and installing the rstat monitor), and mainly monitoring the following network performance indexes aiming at the network performance by using a Linux remote system monitor, wherein the network performance indexes are shown in a table 1:

TABLE 1

Client terminal of testing machine

And performing Socket communication with the virtual machine server end by using network protocols such as TCP (transmission control protocol), UDP (user datagram protocol) and the like, executing a test task, collecting a test result, and sending the test result to the master control machine. The Client supports the testing tasks of three protocols of TCP, UDP and TCP/UDP mixture. The Client can freely select.

Main control machine Manager

Sending a test command to a client test machine, and collecting and counting test results; and sending an RPC request to the virtual machine, acquiring various performance indexes of the virtual machine, and displaying the performance indexes in a chart form.

Collecting and counting distributed test results:

the main control computer collects the test result data of the tester which are collected for times every 3S, averages different collected data, and counts the current Socket transmission rate (M/S).

Monitoring the Linux performance:

linux performance monitoring mainly uses Remote Procedure Call (RPC) technology, RPC is a way of inter-process communication that allows computer programs to execute programs or subroutines of another address spaces (usually another computers on a shared network) without knowing the implementation details of the specific process the most important difference between remote procedure calls and local calls is that remote procedure calls can cause Call failures due to unpredictable network problems, which does not happen with local calls, standardized RPC systems are generated in order for different clients to access the server, most of these systems use Interface Description Language (IDL) to make remote calls available to clients under different platforms through the interface defined by IDL, RPC communication spans the transport and application layers [21], and it is assumed that some transport protocols exist, such as TCP (transport control protocol) or UDP (user Datagram protocol) that carry data between communicating clients, service is served by the client, sending Remote Procedure Calls (RPC) until it sends a request to the server, which is specified in the state of sending a remote procedure Call until the server is completed, , which the server waits for the request as shown in the following figure 361:

(1) the client calls the client handle, which is local procedure calls, and the parameters are passed to the stack;

(2) the client handle encapsulates the parameters into the message and sends the message by calling a system call;

(3) the local operating system of the client sends the message from the local machine to the remote server;

(4) the server operating system receives the arriving message data packet and transmits the data packet to the server handle;

(5) unpacking the message data packet by the server handle, acquiring the transmitted parameters, and calling the server process;

(6) returning the result of the procedure call to the server handle;

(7) the server handle returns a result, and the server operating system packages the result to form a message;

(8) the server operating system transmits the result message back to the client;

(9) unpacking the result message by the client handle, restoring the result data and sending the result data to the client;

(10) and the client receives the data returned by the client handle.

The reference numerals (1) to (10) in fig. 1 correspond to the above-described internal operation steps (1) to (10).

Acquiring real-time parameters of performance monitoring indexes of the virtual machine, and displaying the real-time parameters in a chart form:

the method comprises the steps of firstly defining a data display style by using XSL, then writing test data into an XML file according to a rule determined by , and finally converting the data file in the XML format into a user-friendly Html page through an XSL in a system Xml.

Software installation and use main steps

A Server end:

1. starting an rstat port monitoring program of an operating system; installing the rstat port monitoring tool first requires downloading the rstat's Tar compressed file from the official website. After decompressing the tar file, the rstat monitor can be compiled, configured, and installed.

2. Change the directory to/etc/xinetd.d and create a file named rstatd with:

3. restart the inetd (or xinetd) daemon.

The following commands are used: the method comprises the steps of starting a Socket multithreading monitoring client request program, clicking a Server end monitoring project VirtualNetWork _ Server, and starting a multithreading monitoring program;

client end: and starting a main program of the UI of the client.

Manager main control machine:

and installing and starting the RPC client, configuring ip parameters of the tested virtual machine and various performance detection indexes, and collecting performance monitoring index parameters. The collection step is roughly divided into 4 steps:

1. the RPC client constructs a udp request through a Socket according to the IP address of the target server and the default program port 111 of the portmapper, and sends the request to a portmapper application program of the target machine to inquire the program port number of the rstatd; after receiving the request, the portmapper application queries a mapping table between the port and the application, and returns the port number of the rstatd monitoring program to the RPC client;

2. after the RPC client obtains the port number of the rstatd program, a Socket reconstructs a new Udp request according to the original target IP address and the rstatd at the port number, and sends a request service to the rstatd monitoring program;

3. after the Rstatd monitor receives the service request, the performance data collected by the current performance counter is returned.

And sending the test task to the client.

1. And starting a program of a main control computer, and deploying the main control computer to a tomcat server.

Accessing a main interface of a main control machine Manager to distribute a test task through http:// localhost: 8080/Manager/host.action;

2. and uploading the test task, clicking a test starting button, and distributing the test task to the Client terminal. 3.

3. And displaying various performance parameters of the virtual machine in real time in a chart form.

The network performance indexes mainly comprise Bytes Total/sec, Current Bandwidth, Output queue Lenth, Packets Received/s and packet Received Error.

The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.

Claims

1, VmWare-based virtual machine performance monitoring method, which is characterized by comprising the following steps in sequence:

2. The method for monitoring the performance of the virtual machine according to claim 1, wherein in step S3, the collecting and counting the distributed test results specifically comprises the steps of collecting test result data of the client test machine times every 3S by a main control machine Manager, averaging different collected data, and counting the current Socket transmission rate.

3. The method for monitoring performance of a virtual machine according to claim 2, wherein in step S3, the collecting distributed test results specifically includes the following steps:

a. the RPC client constructs a udp request through a Socket according to the address of the target machine and the default program port 111 of the portmapper, and sends the request to a portmapper application program of the target machine to inquire the program port number of the rstatd;

4. The method for monitoring performance of a virtual machine according to claim 1, wherein in step S4, the obtaining of the performance indexes of the virtual machine and the displaying in the form of a graph specifically includes:

5. The kinds of virtual machine performance monitoring system of claim 1, comprising a virtual machine Server end, a tester Client end, and a host computer Manager connected with each other two by two, wherein the virtual machine Server end starts Socket monitoring in a Server thread pool and monitors a link request of a Client, the tester Client end uses TCP, UDP and other network protocols to perform Socket communication with the virtual machine Server end, execute a test task, collect a test result, and send the test result to the host computer, the host computer Manager sends a test command to the Client tester to collect and count the test result, sends an RPC request to the virtual machine to obtain various performance indexes of the virtual machine, and displays the performance indexes in the form of a chart.