Abstract
Due to the low efficiency of traditional network design, and long cycle of network construction, this paper simulated the campus network model based on the OPNET simulation platform. From the aspects of network nodes, link media, number of servers and settings, etc., the network performance of the campus network solution was improved, and the corresponding solution was verified through the OPNET simulation platform. Simulation experiments has verified the feasibility of network simulation in network planning and design.
Introduction
Network simulation is a new technology, which can verify the network design plan and provide reliable quantitative basis for network planning and design. Through scientific network design and network construction, the utilization rate of the network is improved and the investment risk is reduced. At present, the two most commonly used network simulation tools are NS2 and OPNET. The former is a shared open source code developed by the University of Berkeley. Users can download it directly from the Internet. There are many shortcomings and reliability is not guaranteed due to NS2 is a sharing tool. Users need to model from scratch and apply only to small-scale simulations. The latter is a large-scale communication and computer network simulation software package developed by the United States OPNET Technology Corporation. It supports the object-oriented modeling method and provides a graphical editing interface that is user-friendly. OPNET uses advanced techniques such as time-discrete simulation and hierarchical modeling, it integrates the tools needed in each simulation research phase to form a large-scale simulation system that integrates model design tools, simulation core functions, data collection tools, and data analysis tools. OPNET has become one of the most outstanding ones in the current network simulation software due to the above advantages. Therefore, in this paper the campus network model simulation selects OPNET as a simulation tool.
In foreign countries, especially in Europe and America, the application of simulation technology in network planning and design has become more and more common. Companies such as Cisco and AT&T are all using OPNET to perform various types of simulation tests, controlling the reduction of operating costs, and greatly improving the efficiency of research and development. The United States has widely adopted simulation methods in national defense, urban construction, and other large-scale projects. In China, network planning and design is mainly based on traditional experience, and network simulation in network design has just started.
Leng use OPNET network simulation method of top-down construction [1]; Du et al. use NS-2 simulation software to construct the network model, has carried on the simulation comparison to the link utilization rate, the delay time and other network performance parameter [2]; Shi et al. give the simulation of network delay under different loads [2]; other researchers give the modeling and simulation methods of audio stream, video stream, wireless LAN and so on in OPNET [4, 5, 6]. Gu et al. proposed the establishment of an IP network traffic model based on the OPNET simulation platform; Liu et al. analyzed the modeling mechanism of the IP multicast network in OPNET.The above documents are done at a certain point, and the computer network is a huge and complex system, and many technical indicators restrict and influence each other.
In this paper, starting from the global computer network, various parameters affecting network performance of campus network is analyzed, in the OPNET simulation environment of the campus network is modeled and simulated, simulation results verify the feasibility of network simulation in network planning and design.
Performance evaluation index
Whether the campus network can provide a stable network platform and reliable application services has an important impact on the normal and orderly work of school education, teaching, scientific research and management. To evaluate whether the campus network applications can meet the needs of users, the application response time, network delay, network throughput, network utilization and other performance indicators need to be used [7].
Response time
Response time refers to the time required to send instructions from the user to the network response and start the execution of user instructions, the shorter the response time, the better the performance, the higher the efficiency [8]. The response time of the LAN (Local Area Network) is usually 1
Network delay
Refers to the network delay of a packet from the user’s computer is sent to the web server, and then immediately from the web server to return the user computer round-trip time. Network delay and network delay jitter is small, so the quality of the network is better. The main influence factors of campus network delay is network traffic (network traffic more switch and router queuing time is longer, the network delay is greater). For the video on demand system (Video On Demand, VOD) and other application requirements, the end-to-end delay time should be as small as possible, because of the time delay affect the video quality.
Throughput
Throughput refers to the amount of data (bits, bytes, groups, etc.) that are successfully transmitted in a unit time, such as networks, devices, ports, or other facilities. Usually represented by BPS (bit/sec), Bps (byte/sec) or PPS (packet/sec) [9]. In the campus network environment, in order to ensure the smooth progress of teaching and academic activities, the throughput of the equipment and link must be guaranteed the most basic traffic.
Utilization rate
Link utilization
The link utilization rate in the network refers to the percentage of the network bandwidth occupied by the actual operation. The link utilization rate can’t be too high, otherwise it will cause the delay time of the network to increase. Meanwhile, the utilization of the link is too high, which will cause some services can’t run.
The link utilization refers to the percentage of network bandwidth occupied by the actual running time. The link utilization can’t be too high; otherwise, the network delay will increase. Meanwhile, high link utilization may cause some services fail to operate.
Resource utilization rate
Resource utilization refers to the effective working time of network resources, which accounts for the percentage of the whole work time. Such as channel utilization, memory utilization, CPU utilization and so on. Resource utilization also plays an important role in the performance evaluation index of campus network. By analyzing the utilization rate of each part, the bottleneck position of the network can be effectively judged. In general, the utilization ratio of CPU is best about 50%–40%, and if the utilization rate is too large, the network performance will decline significantly.
Network packet loss
Network packet loss rate refers to the percentage of packets lost during the round-trip process from the client to the server in a specific time interval. Due to the network congestion during the packet switching process, the traffic will be discarded because of network failure. In general, the packet loss rate of the normal network is less than fifteen percent, which will lead to the stagnation of the normal application of the network.
In general, network applications, network applications such as video of this kind, it is to allow a certain degree of packet loss rate, and the dropped packets will not be re sent, but based on network application protocol of network packet loss rate requirement is much higher, if the packet drop, will cause the packet re send additional overhead to send packets will cause the network.
Network simulation process
In the study of network simulation [10], it is generally divided into the following four stages of simulation work to complete the whole simulation process, as shown in Fig. 1.
Simulation flow chart.
The design stage: the establishment of the stage will use the actual needs to complete the simulation scene through simulation modeling, also need to design following experimental series, the first is to set the model input parameters, the appropriate setting simulation statistical content, as well as the simulation program running time, random number seed simulation, simulation running times etc. parameter. Simulation operation stage: in this stage, simulation experiments are carried out using simulation software tools to obtain the later simulation experiment data. Simulation analysis stage: after obtaining the simulation data, the simulation analysis tool and the knowledge of mathematics are used to analyze the simulation results. At the same time in this process can make full use of data filtering technology to choose hoof simulation results for data, as if to the accuracy of simulation data, simulation experiments can be repeated independent operation and simulation results are statistical analysis method using experiments to solve the problem of stochastic network simulation experiment of the maximum. Simulation report phase: this stage is mainly to complete the late Research Report of the network simulation, and write the relevant simulation experiment documents.
A process model for creating an application layer protocol
A process model is actually a finite state machine, which represents the state of the module and its logic (state transfer rules). For the application layer, using the exchange protocol standard, according to each node application layer information exchange between different nodes will be divided into two types: one is only sending message, the message is not received by other nodes; another is sending message, message receiving and other nodes, statistical data received the number of packets, application layer traffic, end-to-end delay and other indicators, and according to the received message reply to the corresponding message type [11]. The application layer uses a unified modeling method to model, so that after building a typical node, team development can be done, and the built model is also easy to maintain.
The process model development is divided into five stages:
The first stage is context definition. Determine the components that are connected to each other; determine the communication mechanism between the components; determine the system and the block diagram of the associated components. The second stage means the analysis process. Determine whether to use multiple processes or single processes to implement the behavior of modeling objects. The third stage, event enumeration (each process). Determine the logical event for each process; determine the event implementation method. In the fourth stage, the event response table (each process) is developed. The fifth stage, the implementation of the process model. The first to fourth stages are the design stage and the fifth stage is the realization stage.
Create node model
The OPNET node model system is a block based approach [12, 13], which includes a series of connection modules. Each module has its own set of I/O, state storage, and internal computing methods. According to the hierarchical structure of TCP/IP network, the application layer application module defines various applications, such as Database, Email, Ftp, Http, etc., they use TCP protocol, such as Video Conferencing, Voice Transport and so on, are using the Conferencing protocol.
Network structure model.
The network model describes the communication system at a higher level, and the communication equipment and communication link together constitute the topology model of the communication system. The network model describes the location of nodes in the network, the links between nodes and the configuration of nodes. The network model can include one or more nodes or subnets, and the size and scope of the network model can also be determined according to the actual design requirements. According to the simulation requirements, each scene in this design is composed of the device nodes with the same priority, which can create the topology of the data network. Network modeling and business modeling are carried out in OPNET. Through simulation, the statistical information parameters of each network performance index are collected, and the network performance is analyzed. According to the network topology, a network simulation model is set up as shown in Fig. 2.
Simulation and analysis
The campus network connects the Internet network through a router, the accuracy of business modeling is the key to evaluate the network performance of the system. The simulation of traffic flow settings must be able to correctly reflect the statistical characteristics of the business, the business is not accurate, will lead to inaccurate simulation results. The campus network defines two applications: application configure and profile configure. In the application configuration, there are six major campus network applications including FTP, HTTP, Database, Print, Voice, and Email. Each application uses two modes, Low Load and High Load. The impact of the size of the business on the performance of the server network can be judged by the collection of different business indicators. The campus network traffic parameters are shown in Table 1.
Business parameters of campus network
Business parameters of campus network
The network node is the network element equipment in the network. In some nodes which need to process a large amount of data, the network equipment is not enough to complete the timely processing of the business, and then the network bottleneck is formed in these positions. According to the flow theorem, the main factors affecting the network performance can be calculated, including these network bottlenecks. To improve the performance of the network, it is an important premise to explore the bottleneck of these networks.
As shown in Fig. 3, the traffic flow of some network nodes is obtained by simulation. In each layer, one nodes are selected in the three layer model, where node_50 is the node of the core layer, node_98 is the distributed layer node, and node_65 is the node of the access layer. Blue represents the flow curve of node_50, green represents the flow curve of node_98, and red represents the flow curve of node_65. What can be seen from the graph is that the traffic of node node_50 is much larger than that of node_98 and node_65 nodes, and it can be inferred that the core layer of the device is the bottleneck of the network, and needs to be improved.
Analysis of network link
The transmission of data depends on the media in the network, and the quality of the link seriously affects the performance of the network, so the impact of the transmission medium on the network is analyzed. By changing the transmission medium, running simulation, their impacts on network performance were tested, so as to find a suitable link design scheme. The core layer adopts the 10 Gigabit fiber link scheme named scenario 1, and the access layer uses the 10Base T twisted pair link scheme named scenario 2, and the simulation results show that Fig. 4 is as follows.
Traffic flow of several network nodes.
Effect of link media on network performance.
Changes of link load in core layer.
The impact of server placement on network device load.
The following conclusions can be drawn from Fig. 3. The enhancement of the transmission quality of the core layer plays an important role in reducing the network delay, so it is necessary to change the core layer link to the high quality medium. However, the link layer changes in the access layer does not cause too much delay change. In order to save the network cost, 10Base T twisted pair is used as the link medium in the access layer.
Before and after the link changes, the load changes in the core layer are shown in Fig. 5.
The blue curve represents the traffic size of the 10Base T twisted pair network, and the red curve represents the flow size of the 10 Gigabit fiber network. The information obtained from the graph is that the high quality transmission link can accelerate the speed of data processing, reduce the possibility of network congestion, and also reduce the network traffic.
Therefore, the improvement of the core layer link is to improve the network performance directly, so as to get an effective method of cost-effective design scheme. Therefore, in the final choice of the link, the gigabit optical fiber link is used to connect the core layer equipment, and the 10Base T twisted pair link is used to connect the access layer device, taking into account the cost at the same time, to maximize the performance of the network.
In addition to a very small number of information in the network, the server needs to provide support, after the interaction and then the destination of the transmission and processing, so the rationality of server configuration will also affect the performance of the network.
Mainly from the location and number of servers, the impact of network performance is analyzed, the server is configured to the core layer, named Scenario1, and then configured to the distribution layer, named Scenario2, through the operation simulation, get the graph 5 of the core layer equipment load graph.
In Fig. 6, the red and blue curves represent the core layer device load curve when the server is placed on the distribution layer and the core layer respectively. The results show that after the server is placed in the core layer, the amount of traffic that needs to be handled is increasing dramatically, the core routing load is aggravated, and the network performance is greatly decreased. Therefore, the server needs to be placed in the access layer to reduce the traffic flow of the core layer, avoid the overload of network equipment, and affect the network performance of the core layer.
The number of access layer servers expanded to two times, the original network named Scenario2, the number of servers increased network default Scenario1, the operation simulation to get the following network delay curve Fig. 7.
The influence of server number on network delay.
It can be seen that after the server increases, the network data processing ability has improved a lot, the network delay has been reduced by nearly 1/5. Therefore, in the network design and planning, if there are sufficient funds, you can consider increasing the number of servers in the access layer, to improve network performance.
The campus network model was established using a top-down hierarchical campus network design method. Through the analysis of network nodes, network links, and server configurations, the campus network performance was effectively improved. It was concluded that the core layer uses high performance Cisco7600 series routing equipment to solve the network bottleneck. At the same time, it laid the transmission link of ten thousand megabytes, configuring the appropriate number of servers to improve the network performance, using the transmission link of the 10Base T twisted pair to lower the network cost to the access layer, thus an efficient and reliable campus network design scheme was obtained.
Conclusions and further work
This paper discussed the factors that affect the performance of the campus network, introduced the principles of network simulation and workflow, and used OPNET network simulation software to compare and analyze network nodes, network links, and server configurations to identify network bottlenecks. At the same time, new services were added to the network to examine changes in network performance, which provided objective quantitative basis for network upgrades and ensured the reliability of decision-making in the process of network upgrade and reconstruction. Combining with the actual needs of the campus network and the popular campus network technology, a reform program was proposed. The simulation results showed that the new scheme was feasible, and it also indicated that OPNET was an effective tool for network planning and design.
This paper did not analyze the network performance of the widely used video streaming data in the campus network. The wireless network is an important direction for the future development of the campus network, especially in many campus network planning nowadays, the wireless network has been added, however the paper did not involving this part of the content for discussion. The next work will be focused on further research based on the shortcomings of existing papers: they will mainly be focused on the use of video streaming technology in campus networks and simulation of wireless LANs, etc.
Footnotes
Acknowledgments
This study was supported by 2016 Yangtzeu University Teacher Development Special Project Fund (Project ID: JS201704).