CN110932879B

CN110932879B - Network performance test method, data processing device and storage medium

Info

Publication number: CN110932879B
Application number: CN201811101209.1A
Authority: CN
Inventors: 袁诚
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Cloud Computing Ltd
Priority date: 2018-09-20
Filing date: 2018-09-20
Publication date: 2022-12-23
Anticipated expiration: 2038-09-20
Also published as: CN110932879A

Abstract

The invention discloses a network performance test method, data processing equipment and a storage medium, wherein the method comprises the following steps: establishing a plurality of connections at least one client of a network, wherein each connection corresponds to a channel of the network, and each channel corresponds to a remote address of the network; when the at least one client generates data traffic, distributing the generated data traffic to the corresponding channel according to a preset distribution strategy; and analyzing the corresponding network performance based on the data traffic processing process of the channel. The embodiment of the invention can realize network performance test.

Description

Network performance test method, data processing device and storage medium

Technical Field

The present invention relates to the field of computer technologies, and in particular, to a method and an apparatus for testing network performance, a data processing device, and a computer-readable storage medium.

Background

In the past, people simulate network traffic through a network traffic tester (such as tesgine, ixia), and the method has many defects: 1. the tester is a physical hardware device, and the purchase investment can cause high cost; 2. the port of the receiving and sending packet of the tester equipment is fixed, and the large-scale cluster environment cannot be butted; 3. the flow model configuration of the tester equipment is complex; 4. the tester only can simulate the flow of a seven-layer network, and focuses on testing the forwarding performance of the network equipment.

However, for large-scale cluster environments, hardware devices such as network traffic testers are no longer suitable. First, the tester device cannot simulate various pressures (CPU, memory, network card, interrupt, etc.) of network traffic on the physical machine in the large-scale cluster environment, and also cannot simulate the influence of various thread models on Remote Direct Memory Access (RDMA) network performance on the physical machine in the large-scale cluster environment.

Disclosure of Invention

In view of this, the present invention provides a method and an apparatus for testing network performance, a data processing device, and a computer readable storage medium, which can implement network performance testing and solve the problem of difficulty in testing network performance in a large-scale cluster environment.

In a first aspect, the present invention provides a network performance testing method, where the method includes: establishing a plurality of connections at least one client of a network, wherein each connection corresponds to a channel of the network, and each channel corresponds to a remote address of the network; when the at least one client generates data traffic, distributing the generated data traffic to the corresponding channel according to a preset distribution strategy; and analyzing the corresponding network performance based on the data traffic processing process of the channel.

According to the method of the present invention, the analyzing the corresponding network performance in the channel-based data traffic processing process includes: each channel sends the data packet to the corresponding remote address based on the divided data flow; collecting network pressure parameters in the process of sending data packets by each channel; and after the data packets of all the channels are sent, analyzing the collected network pressure parameters to obtain a test result.

According to the method of the present invention, the establishing a plurality of connections at least one client of a network comprises: establishing a plurality of threads at the at least one client, and establishing at least one connection in each thread.

According to the method, each thread distributes the data traffic of each corresponding connection in a token mode.

According to the method, each client distributes the data flow of each corresponding thread according to a preset distribution strategy.

According to the method of the present invention, the distribution policy includes at least one of the following policies: poll Queue Pair (QP), random QP, fixed QP.

According to the method of the present invention, before at least one client of the network establishes a plurality of connections, the method further comprises: and generating configuration information of the at least one client according to the input information.

According to the method of the invention, the configuration information includes a list of hostnames and/or a list of internet protocol IP addresses of all clients.

The method according to the invention, further comprising: and deploying the at least one client to the network according to the configuration information of the at least one client.

In a second aspect, the present invention further provides a network performance testing apparatus, including: the system comprises a connection establishing module, a network processing module and a network processing module, wherein the connection establishing module is used for establishing a plurality of connections at least one client of a network, each connection corresponds to one channel of the network, and each channel corresponds to a remote address of the network; the flow distribution module is used for distributing the generated data flow to the corresponding channel according to a preset distribution strategy when the at least one client generates the data flow; and the network performance analysis module is used for analyzing the corresponding network performance based on the data traffic processing process of the channel.

In a third aspect, the present invention also provides a data processing apparatus, comprising:

a memory for storing a program;

a processor for executing the program stored in the memory to perform the method as described above.

In a fourth aspect, the present invention also provides a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method as described above.

The invention can establish a plurality of channel connections in the network, and when the client generates the flow, the flow is distributed to the channel connections according to the distribution strategy, and the data packets are sent in the channel connections, thereby simulating the sending process of the data packets, obtaining the corresponding network performance parameters, and realizing the test of the network performance.

Drawings

Fig. 1 is a flowchart of a network performance testing method according to an embodiment of the present invention.

Fig. 2 is a flowchart of a method for testing performance of a cluster network according to an embodiment of the present invention.

Fig. 3 is an architecture diagram of a pressure client module according to an embodiment of the present invention.

Fig. 4 is a schematic diagram of a hardware structure of a data processing apparatus according to an embodiment of the present invention.

Detailed Description

The invention is described in detail below with reference to the drawings and examples. It should be understood that the specific embodiments described are merely illustrative of the invention and are not intended to limit the invention. Terms such as first, second, etc. in this document are only used for distinguishing one entity (or operation) from another entity (or operation), and do not indicate any relationship or order between these entities (or operations); in addition, terms such as upper, lower, left, right, front, rear, and the like in the text denote directions or orientations, and only relative directions or orientations, not absolute directions or orientations. Without additional limitation, an element defined by the phrase "comprising" does not exclude the presence of other elements in a process, method, article, or apparatus that comprises the element.

Remote Direct Memory Access (RDMA) can be regarded as a Remote DMA technology, and RDMA is a technology for solving a delay of server-side data processing in network transmission. RDMA allows user-mode applications to read or write directly to remote memory without kernel intervention and memory copying occurring. Initially, RDMA technology was only applied in the area of high performance computing; currently, RDMA technology is gradually coming into the field of view of more and more developers due to the increasingly prominent network bottlenecks in large-scale distributed systems and data centers.

RDMA network performance observation in a large-scale cluster environment has many complex scenes, such as one-to-one, one-to-many, many-to-many, and the like, and traffic models are also complex, such as burst traffic, large network packets, small network packets, large and small packet hybrid packets, and the like, and in addition, indexes to be observed are also many, such as throughput, delay, stability, whole cluster forwarding performance, single machine forwarding performance, and the like.

Based on the situation, the invention provides a network performance test method, and the performance test of various networks can be realized by utilizing the network performance test method. Particularly in a large-scale cluster environment, the method can conveniently and quickly obtain a plurality of items of data in the network in the large-scale cluster environment for analyzing the test result. Various aspects of the invention are described in detail below.

Fig. 1 shows a flow chart of a network performance testing method according to an embodiment of the present invention, which includes:

s101, establishing a plurality of connections at least one client of a network, wherein each connection corresponds to one channel of the network, and each channel corresponds to a remote address of the network;

s102, when the at least one client generates data traffic, distributing the generated data traffic to a corresponding channel according to a preset distribution strategy;

and S103, analyzing the corresponding network performance based on the data traffic processing process of the channel.

By using the scheme, a plurality of connections can be established in the network, and when the client generates the flow, the flow is distributed into the connections according to the distribution strategy, and the data packets are sent in the connections, so that the network performance in the sending process is simulated and obtained.

In one embodiment of the present invention, the client establishes a plurality of threads, and then establishes at least one connection in each thread, thereby obtaining a plurality of connections. The client distributes the flow to a plurality of threads, and then the flow is distributed to a plurality of respective connections by the threads.

Further, each connection channel sends the data packet to the corresponding remote address based on the divided data flow; collecting network pressure parameters in the process of sending data packets by each channel; and after the data packet of each channel is sent, analyzing the collected network pressure parameters to obtain a test result.

Therefore, the invention can provide flexible and changeable scene simulation capability by establishing a multi-stage distribution processing mechanism and a flow model abstract processing mechanism in the network.

In an application scenario of the present invention, the performance of the RDMA network in a large-scale cluster environment is simulated and tested, and the processing procedure of the method of the present invention can be implemented by the following embodiments.

The following modules were constructed in the present invention: the system comprises a deployment module, a configuration module, a pressure client module, a pressure server module and a data analysis module, wherein the deployment module specifically comprises:

● A deployment module: the system is responsible for rapidly deploying a configuration module, a pressure client module and a pressure server module to a cluster environment;

● A configuration module: the system is responsible for generating the configuration of the pressure client module and the pressure server module according to the input of a user;

● A pressure client module: the system is responsible for generating network flow pressure and receiving a response request of a pressure server module;

● The pressure server side module: the system is responsible for receiving network flow pressure and responding to the request of the pressure client module;

● A data analysis module: and the system is responsible for analyzing and counting the test results.

Figure 2 shows a flow chart of a method for testing the performance of a cluster network according to one embodiment of the invention,

fig. 3 shows an architecture diagram of a pressure client module of an embodiment of the present invention. Wherein, pressure client module includes: the main thread Monitor, the thread Worker, the Channel and the distribution Policy specifically:

● Channel: a channel represents an RDMA QP, corresponding to a particular remote address. Wherein QP is an abbreviation of Queue-Pair, representing a reliable connection in RDMA networks, and the concept of peer-to-peer in the TCP/IP protocol suite is socket.

● And (2) Worker: one Worker represents one thread, and manages the traffic quota of each Channel in the Worker in a token mode.

● Monitor: and the main thread is responsible for managing the flow quota of each Worker.

● And Policy: a distribution strategy is used for controlling the number of data packets sent by which ones of a group of workers or a group of channels each time the client generates flow; for example: polling QP, random QP, and/or fixed QP.

For the multi-level distribution mechanism and the traffic model abstraction mechanism established by the invention, the multi-level distribution mechanism is a hierarchical mechanism of Monitor → Worker → Channel to realize the distribution of thread level and network connection level, so that various complex distribution models can be constructed. The traffic model abstraction mechanism abstracts large-scale cluster network traffic into two layers of single-connection traffic attribute and connection distribution model.

The single connection traffic attributes comprise traffic attributes in traditional network communication, such as packet size, pressure depth and the like; the connection distribution model defines how traffic is distributed on each connection, and basic distribution strategies include random, polling, fixed and the like; the connection distribution model describes the distribution of network traffic in a large-scale cluster.

The invention can provide flexible and changeable scene simulation capability through the multi-level distribution model and the flow model abstraction; the invention has simple configuration, can realize batch push and greatly simplifies the operation complexity of users; the invention breaks through the scale limitation of hardware through a software implementation mode, not only expands the application range and is easy to realize, but also can reduce the input cost to a certain extent.

The following describes an alternative specific process of the embodiment of the present invention by way of a specific example. It should be noted that the scheme of the present invention does not depend on a specific algorithm, and in practical applications, any known or unknown hardware, software, algorithm, program, or any combination thereof may be used to implement the scheme of the present invention, and the scheme of the present invention is within the protection scope of the present invention as long as the essential idea of the scheme of the present invention is adopted.

In one embodiment of the present invention, consider a scenario in which 60000 connections are established at a single client in a cluster of 60 computers; and testing the network delay in the scene that each connection in the cluster sends the network packets in sequence.

It should be noted that the practical application scenarios corresponding to the test model are as follows: in a large-scale cluster, a large number of users have sporadic stress, and in this scenario, delay loss caused by the hopping of traffic on different connections should be concerned.

i. Preparation configuration: in the embodiment, a client is adopted to establish 60 workers, and each Worker establishes 1000 channels; each server has 10 ports, and the mode of 60 servers is total; then 100 connections will be established per port per server.

The ServerList comprises a list of hostname or ip of 60 machines;

the ServerPortCount =10, each pressure server has 10 ports, and each port can be connected with 100 clients;

the ClientList comprises a hostname or an ip of a machine which initiates traffic;

where Depth =1, is the delay under test single Depth pressure;

policy = number of threads of polling client =60, and number of channels owned by each thread of client =1000;

wherein, the MsgSize and the like are optional configurations and can be set according to different requirements of users;

deployment testing: the configuration module translates the user configuration into json format and pushes the json format and the json format to a corresponding machine together with the pressure tool;

start of the test: the pressure client side is started, a Monitor and 60 workers are established, 1000 channels are established in each Worker, 60000 channels are calculated in total, and the 60000 channels are connected with 600 far-end addresses;

then, monitor applies polling Policy to distribute quota to each Worker in turn, and Worker applies polling Policy to distribute quota to each Channel below in turn.

And after the Channel obtains the quota, sending the data packet.

So as to reciprocate. Here, since Depth =1, that is, the global total quota is 1, that is, only one Channel has a sending right at the same time, in combination with Worker polling, a distribution mechanism of Channel polling simulates a scenario in which 60000 connections send data packets in sequence.

Wait for the end of the test, collect and analyze the data.

The embodiment of the invention is utilized to test the network performance, such as the performance of a large-scale cluster network, and has the following advantages:

(1) the invention is realized by a software mode, has low cost and quick deployment.

(2) The invention provides a convenient configuration mode (such as hostname list or IP list), theoretically supports cluster environments of any scale, and has no limit on the number of the packet receiving and transmitting ports.

(3) The invention provides a common preset flow model and a structural configuration method, and the configuration is simple.

(4) The running environment of the invention is the cluster physical machine, therefore, the invention naturally has the capacity of occupying hardware resources such as CPU, memory, network card, interrupt and the like besides only simulating network flow, and can self-define various thread models and flow models.

(5) In addition, the invention provides flexible flow model and thread model setting, and can carry out combined simulation on various complex pressure scenes.

(6) The invention can be applied to not only a distributed storage system, but also any other fields using RDMA technology, such as network, large-scale computation, and the like.

In addition, the invention also provides a device, a data processing device and a computer storage medium corresponding to the method of the embodiment of the invention.

Wherein the data processing apparatus comprises:

a memory for storing a program;

a processor configured to execute the program stored in the memory to perform the steps of the method according to the embodiment of the present invention.

The present invention also provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the computer program instructions implement the steps in the network performance testing method according to the embodiment of the present invention, including:

establishing a plurality of connections at least one client of a network, wherein each connection corresponds to a channel of the network, and each channel corresponds to a remote address of the network;

when the at least one client generates data traffic, distributing the generated data traffic to the corresponding channel according to a preset distribution strategy;

and analyzing the corresponding network performance based on the data traffic processing process of the channel.

It should be noted that in the above embodiments, all or part may be implemented by software, hardware, firmware, or any combination thereof. When used in whole or in part, is implemented in the form of a computer program product that includes one or more computer program instructions. When loaded or executed on a computer, cause the flow or functions according to embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer program instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer program instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center via a wired (e.g., coaxial cable, fiber optic, digital Subscriber Line (DSL), or wireless (e.g., infrared, wireless, microwave, etc.) link. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that includes one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid State Disk (SSD)), among others.

Fig. 4 is a block diagram illustrating an exemplary hardware architecture capable of implementing methods and apparatus according to embodiments of the present invention, such as apparatus of embodiments of the present invention. Computing device 1000 includes, among other things, input device 1001, input interface 1002, processor 1003, memory 1004, output interface 1005, and output device 1006.

The input interface 1002, the processor 1003, the memory 1004, and the output interface 1005 are connected to each other via a bus 1010, and the input device 1001 and the output device 1006 are connected to the bus 1010 via the input interface 1002 and the output interface 1005, respectively, and further connected to other components of the computing device 1000.

Specifically, the input device 1001 receives input information from the outside and transmits the input information to the processor 1003 via the input interface 1002; the processor 1003 processes the input information based on computer-executable instructions stored in the memory 1004 to generate output information, stores the output information temporarily or permanently in the memory 1004, and then transmits the output information to the output device 1006 through the output interface 1005; output device 1006 outputs the output information external to computing device 1000 for use by a user.

The computing device 1000 may perform the steps of the network performance testing method of the present invention, including:

Processor 1003 may be one or more Central Processing Units (CPUs). When the processor 601 or the processor 701 is one CPU, the CPU may be a single-core CPU or a multi-core CPU.

The memory 1004 may be, but is not limited to, one or more of Random Access Memory (RAM), read Only Memory (ROM), erasable Programmable Read Only Memory (EPROM), compact disc read only memory (CD-ROM), a hard disk, and the like. The memory 1004 is used to store program codes. It is understood that the functions of any module or all modules provided in the embodiments of the present invention may be implemented by the central processing unit 1003.

It should be noted that, all parts of this specification are described in a progressive manner, and the same and similar parts among the various embodiments are referred to each other, and each embodiment is described with emphasis on being different from the other embodiments. In particular, for apparatus and system embodiments, the description is relatively simple as it is substantially similar to the method embodiments, and reference may be made to the description of the method embodiments in this section for their relevance.

Claims

1. A method for testing network performance, the method comprising:

establishing a plurality of connections at least one client of a network, wherein each connection corresponds to a channel of the network, each channel corresponds to a remote address of the network, and the plurality of connections comprise at least one connection for each of a plurality of threads of the at least one client;

2. The method of claim 1, wherein the channel-based data traffic process analyzing the corresponding network performance comprises:

each channel sends the data packet to the corresponding remote address based on the divided data flow;

collecting network pressure parameters in the process of sending data packets by each channel;

and after the data packets of each channel are sent, analyzing the collected network pressure parameters to obtain a test result.

3. The method of claim 1, wherein establishing a plurality of connections by at least one client of the network comprises:

establishing a plurality of threads at the at least one client, and establishing at least one connection in each thread.

4. The method of claim 3, wherein each thread distributes data traffic for each respective connection by way of tokens.

5. The method according to claim 3, wherein each client allocates data traffic of each corresponding thread according to a preset distribution policy.

6. The method of claim 1, wherein the distribution policy comprises at least one of the following policies: poll Queue Pair (QP), random QP, fixed QP.

7. The method of claim 1, wherein before the at least one client of the network establishes the plurality of connections, the method further comprises: and generating configuration information of the at least one client according to the input information.

8. The method of claim 7, wherein the configuration information comprises a list of hostnames and/or a list of Internet Protocol (IP) addresses of all clients.

9. The method of claim 7, further comprising: and deploying the at least one client to the network according to the configuration information of the at least one client.

10. A network performance testing apparatus, the apparatus comprising:

a connection establishing module, configured to establish a plurality of connections at least one client of a network, where each connection corresponds to a channel of the network, each channel corresponds to a remote address of the network, and the plurality of connections include at least one connection for each thread of a plurality of threads of the at least one client;

the flow distribution module is used for distributing the generated data flow to the corresponding channel according to a preset distribution strategy when the at least one client generates the data flow;

and the network performance analysis module is used for analyzing the corresponding network performance based on the data traffic processing process of the channel.

11. A data processing apparatus, characterized in that it comprises:

a memory for storing a program;

a processor for executing the program stored in the memory to perform the method of any one of claims 1 to 9.

12. A computer-readable storage medium on which computer program instructions are stored, which when executed by a processor implement the method of any one of claims 1 to 9.