Access Network Performance Optimization in five steps

The Optimization of the access network performance is an important part of the daily work of network administrators. The improvement of the access network performance can bring a better network environment for users. The main goal of the network administrator is to ensure the stability of the access network performance. Although the goal is relatively simple, it is not easy to do it. Based on my experience, I believe that the following five tasks must be done in a down-to-earth manner to improve the access network performance from the root cause.

Task 1: Collect network application baselines

Different enterprises have different network applications, so they have different network performance requirements. For example, if an enterprise has multimedia applications such as video conferencing, there will be strict requirements on the access network performance. Therefore, when improving the access network performance, the network administrator must have a goal, that is, what kind of access network performance is acceptable to the Enterprise.

Generally, if enterprise network deployment is not complex, you only need to collect the CPU, memory, buffer management, Link Utilization, throughput, and other parameters of the vro or vswitch. However, if an enterprise has some special performance requirements, it is best to collect other baseline data except those above parameters, such as availability baselines and network congestion baselines. These baselines are like a reference indicator, which can be used to specify a direction for subsequent work. For example, you can compare the network environment before and after adjustment to determine whether the expected target is achieved.

In fact, for the access network performance, only better, not the best. Therefore, as a network administrator, there is no need to pursue optimal performance. Generally, as long as the current network performance can meet the normal application requirements of enterprises. Therefore, it is of great guiding significance to collect the baseline of network applications for performance optimization.

Task 2: Perform hypothesis Analysis on network-related applications

When building a network or upgrading a network, you cannot only consider the current application when considering the load and performance. You must also consider the network applications that may be used in the future. In addition, the hypothesis analysis includes solution modeling and verification. For example, when adding an application to an existing network, you need to analyze the impact of the application on the existing network performance in advance. The network administrator should record these influences in writing, including changes in the network architecture, main network device parameters (CPU and memory usage), and configuration files. This not only optimizes the network performance in the future, but also provides good data support to solve the resulting network faults.

When a new application is added, leading to performance bottlenecks or network faults, you can use this information to quickly find the cause of the problem and solve it. However, as far as I know, many network administrators are eager to do this. Before they adjust existing network applications, such as adding enterprise antivirus software or adding a file server, they will not make assumptions about these applications. Let's talk about it later. Wait for the problem to be solved. This will undoubtedly increase the time to solve the problem and sometimes increase unnecessary costs. I am very opposed to this practice.

Task 3: Determine the network overhead of the currently used applications

The network application overhead of an enterprise can be divided into two categories: overhead of the network management service and overhead of the non-network management service. Sometimes, although it is necessary to frequently enable the network management service, it also occupies a large amount of network bandwidth, thus affecting network performance. For example, SNMP polling of network devices will consume the CPU cycle of the switch or vro. For example, remote database backup will occupy the network bandwidth and database server resources. As a network administrator, you must be aware of the adverse impact of these management services on the performance of the existing network environment. In other words, you need to know two aspects. First, you need to know that if the current network conditions are not ideal, that is, the relevant indicators have reached the highest level, you can run those management services. At this time, some management services may not occupy a lot of resources. In this case, everyone can run the services for management purposes. However, some management services, such as file backup, occupy a large amount of resources and should be suspended temporarily. Second, we need to reasonably arrange the time for various management services. For example, you can separate the management service from the conventional application service. The most common is to run software upgrade, file backup, server antivirus and other management services when employees get off work and the network is not busy. Minimize the negative impact of management services on network performance.

Task 4: regularly view the network and application capacity information

In my practical work, I will regularly view the network and application capacity information to understand the network operation status. If some enterprises may have ERP and other related application systems, their utilization rates generally have certain rules. For example, ERP and other systems will be used frequently in the afternoon. In a week, the application system related to Monday is the most frequently used, especially the mail system. Because no email is received during the two days of the weekend, the mail capacity will increase significantly when I first arrive at work on Monday. Network administrators sometimes need to check the capacity to understand the network peak periods. Understanding the capacity information and other parameters can help the network administrator determine the performance bottleneck of the Enterprise access network, so that the administrator can take corresponding measures to solve the bottleneck.

In general, in order to make the obtained data more reasonable, the network administrator must repeat the data collection and data analysis work on a regular basis. However, from Task 3, we can see that the Management Service will affect the access network performance. Therefore, how long is the repetition cycle, which is also a critical issue. Not enough. In general, there are no strict requirements for this repetition cycle, mainly depending on the network applications of enterprises. The author's suggestion is, in weeks. Because many enterprises are on vacation on weekends. In this case, a week is a reasonable cycle. In particular, Monday morning may be the bottleneck of enterprise network performance.

Task 5: Make plans in advance to reduce the downtime

The most important purpose of these tasks is to adjust the network configuration and improve the access network performance. However, a very important issue needs to be paid attention to when it comes to improving the access network performance. That is, when optimizing the network configuration, you also need to consider network planning in advance to reduce the time needed to run the machine. To achieve this, the most important step is to establish an advanced process to deal with future capacity requirements. Simply put, you cannot only consider the current needs when optimizing network configurations. If conditions permit, it is best to perform advanced configuration as appropriate.

For example, enterprises may need 10 M Internet bandwidth based on their existing needs, while enterprises only have 4 M. In this case, you need to increase bandwidth or adjust the priority of network traffic to improve performance. However, in the future, enterprises may plan to use network applications such as video conferencing. At this time, the future needs must be considered as long as the enterprise cost permits. For example, enterprises can convert Internet access to optical fiber, rather than dial-up. This will increase the related costs, but will reduce the trouble of subsequent upgrades and shorten the time for users to work on the machine. After all, for some applications, once the machine is used, it may cause great losses to the enterprise. In short, network performance optimization is a complex process. Or, it's not a brave war. The above five tasks are the basic and core tasks in this job. The author details, this will bring some guiding significance to everyone's performance optimization work, reduce the obstacles in the network optimization project.