How to optimize overseas game operations and build a stable global network

This article is the on-site dry goods of WOT2016 Internet Operation and Developer Conference. The new session with the theme of WOT2016 Enterprise Security Technology Summit will be held at JW Marriott Hotel in Beijing Pearl River Delta from June 24th to 25th, 2016!

In the process of development over the years, a large number of corporate users, along with the development of the industry and the entire environment, have gradually seen the economic development speed and network maturity of Southeast Asia, as well as the large overseas demand of many domestic industries. Therefore, Anchang began to expand overseas in 2013.

Basic introduction of Anchang

Anchang has always adhered to the concept that users' businesses are always inseparable from rich, efficient, secure and stable networks, and has spent a lot of time monitoring the network, whether it is network security, chain load, or many problems of the network itself. Liu Xin said that Anchang hopes to not only provide customers with basic networks, but also provide some support for customers' businesses.

What does "rich" mean? The domestic Internet is mainly provided by relatively stable operators such as China Telecom, China Unicom, and China Mobile. However, when domestic enterprises and cloud computing companies enter overseas markets, they will face very complex networks. Focused users will choose cloud computing giants such as Amazon or Softlayer because they can quickly deploy services and meet business needs. For some cloud computing enterprises or large Internet companies, first of all, cloud computing companies cannot meet the requirements of these companies, and when their own development needs to cover more industries, international operators become the first choice, such as Pacific Telecom, PCCW, and NTT.

Its advantages are that the network is standardized and has a relatively comprehensive coverage, and it can be used directly without having to bear too much pressure on the network's own operation and maintenance. The network cost is also low, unlike most domestic cloud computing companies that open nodes abroad.

As for the disadvantages, it is also because of the lack of operation and maintenance capabilities and the network is too standard, so activities are limited and there is no way to flexibly optimize or improve network quality according to business development. For example, international operators will spend a lot of time on deployment when users have certain business needs, but the best time for business has passed. The related investment, risks, and operation and maintenance pressure will be very high.

For Anchang, the first step is to select rich resources, including some international operators. Only by accessing sufficient resources can we provide users with the most basic network needs, such as Pacific Telecom, PCCW, China Telecom Global, China Unicom Global, etc.

Anchang's philosophy is to take the best of all and apply precise routing technology on the international network. The implementation method is to monitor the links and networks of each operator in detail. There are many operators in the local area, not three, but five. How can we ensure that users can smoothly connect to the game and experience the game no matter which operator they use?

Anchang first established a precise monitoring system, from the main network to different countries and regions, and the delay and data transmission of different links were precisely monitored. Then, based on the output of network information, the differences between different link operators in the same region were synchronously compared. For example, when going to Japan through different operators or different links, which one has the best delay and which one has the most stable data transmission? These are extracted from a real-time monitoring library. Finally, through the monitoring data and the model rules of the network, the best and suboptimal link strategies to reach different countries, regions, or different operators in the same region are manually screened out. Try to achieve the best possible delay and the best data transmission quality.

After three years of optimization, Anchang has optimized Hong Kong, Japan, South Korea, Europe, and North America. It has finally developed more than 300 effective strategies for operators in different regions, and the average global latency can be reduced by three to five milliseconds. Previously, the latency from Hong Kong to Germany was 328 milliseconds. After selecting and optimizing the link, the latency is 273 milliseconds. Although it is still based on an ordinary wide area network, it is much better than before.

What is considered efficient? Efficiency means providing higher data transmission efficiency under stable connection and network conditions to ensure that the business can run well. In this process, Anchang explored many ways and finally adopted WAN optimization technology and integrated it into the WAN Internet. This is actually not a very new or particularly technical technology, and is generally used more in CDN. However, Anchang upgraded this technology and integrated it into the regular IDC and cloud computing process as a basic feature of the network.

Under normal circumstances, due to the problems of the TCP protocol itself, when packet loss and delay are detected, it will automatically assume that the quality of the link will decrease, and will automatically adjust the sliding window to reduce its transmission efficiency. In the process of perception, when packet loss is detected on the link, the efficiency of data transmission will drop sharply. When the delay increases, data transmission will also decrease. However, this TCP judgment mechanism is not so perfect, so Anchang changed the traditional TCP acceleration method. The traditional method uses the packet loss principle or the delay principle to judge, and in this way, it manually specifies or manually improves its data transmission efficiency and swings the sliding window.

This is a comparison of Anchang in different environments.

After changing the network link to make it more stable and making data transmission faster, security issues will arise. How should service providers share security issues with customers? Anchang will face some common security threats, with few types of threat identification, small numbers, and large threat vulnerabilities; the threat identification speed will be slow, slow to take effect, and low efficiency. For general Ddos attacks, whether it is equipment or suppliers, the identification speed is tens of seconds, minutes, or even longer. However, for an enterprise or for a customer, when network resources are not so abundant or sufficient, a speed of tens of seconds is enough to paralyze its network. With more and more attacks, the attacks on the network are also evolving, and more tend to target businesses, network weaknesses, and even architectures. These are also problems faced by many companies, but it is difficult for conventional manufacturers to prevent them.

Here Liu Xin cited a case of Anchang based on Huawei.

1. Local cleaning

If the attack traffic is smaller than the local protection capability and will not cause network uplink congestion, the local protection system will perform local protection to protect the attacked network.

Local cleaning is mainly aimed at small-flow, application-based attacks, and mainly defends against C&C and SYN flood attacks.

2. Operator Cleaning

If the attack traffic is larger than the local cleaning capacity or has caused network uplink congestion, the local protection system sends a request to the operator's cleaning center for traffic cleaning.

At the same time, the traffic of the attacked network (/24 bits) is switched to the operator as a whole through BGP. (International and domestic inbound traffic are switched to the backup international lines respectively).

The operator’s cleaning service serves as a second barrier.

3. Cloud cleaning and re-injection

Huawei CCC notified the Global Cloud Cleaning Center to release diversion routes to divert traffic destined for the attacked address to the cleaning center for cleaning.

Anycast diversion: The cloud cleaning center publishes diversion routes within the area it protects, and diverts traffic whose destination address in the area is the customer's attacked network segment address to the cleaning center.

After cleaning at the cleaning center, the cleaned traffic is injected back into the network through the tunnel/backbone network.

Do richness, efficiency, and security really mean stability? Liu Xin shared some of the evolution of Anchang Network.

The first is the overseas network built by Anchang through a three-layer network . It has built a BGP network called FastFiber, which starts from China, goes to Southeast Asia and Asia Pacific, and covers North America, providing network coverage in multiple regions. It consists of three parts: the first is the wide area network; the second is the FastFiber backbone network, a physical network built based on transmission and physical links; the third is an MPLS network built based on the FastFiber backbone network. The interaction, activation and maintenance methods of the MPLS network are simpler and more suitable for users to use, rather than for enterprises to maintain themselves.

The purpose of Anchang's construction of NetworkPOP is to enable each network node to establish network exchange based on the wide area network and provide richer network access.

The second is to maintain an important business , that is, business separation. Separate the important protocols or businesses of customers and connect them through better links.

The third is WAN optimization , which basically includes link optimization, protocol optimization, separation of important network services and application acceleration of services. In addition to conventional networks, Anchang also provides customers with some network features. The purpose is to better fit the customer's business and make the business more stable and secure.

FiberPOP is a highway that connects data centers. Its basic capabilities are important bridges for cross-regional and network interconnection. By opening it up, customers can reduce their own architectural pressure during business deployment.

Net Pop: Basic network coverage and network optimization

IP Backbone Network: Advanced support for networks in Europe, North America, Taiwan, and Singapore

Fiber POP & Fiber: important network carriers with cross-regional, low latency, high transmission and multiple access

Features: Supports flexible network delivery, elastic expansion, and has rich-level network scheduling capabilities.

Introduction of Mpls

Stability is the starting point without an end for network, cloud computing or IC companies. Anchang has been advancing in the process of exploration, improving in the process of practice, growing together with customers, and understanding their business and needs. Anchang's goal is to provide the most basic guarantee for customers' business and help the development of their business.

This article is compiled from the wonderful speech by Anchang Product Director Liu Xin entitled "Game Operations and Global Network Optimization and Management" at the WOT2016 Internet Operation and Developer Conference hosted by 51CTO Media.

Lecture video: http://edu..com/lesson/id-100751.html

Lecturer Profile:

[[167175]]

Liu Xin , Product Director of Anchang Overseas Business Department. Since joining Anchang, he has been responsible for the IDC product line and overseas business related business, ensuring the stability, security and reliability of the data center and core backbone network, and is committed to providing good operation and maintenance and network support for the rapid development of user business.