CDN Accelerated Practice of Huosan CDN in Large File Download and Video on Demand

1、 Content distribution challenges for large file downloads and video on demand

For download type scenarios, common segmentation scenarios include: mobile app stores, game app stores, game updates, and mobile operating system updates. For video type application scenarios, the main segmentation scenarios include short videos, long videos, online education, OTT set-top boxes, radio and television media, and music websites. Why would you compare downloaded application scenarios with video on demand application scenarios? Because the application scenario of video on demand is actually a large file download scenario, the two application scenarios have high business similarities and many common issues. The difference between them is that video on demand has higher requirements for access performance.

The challenges faced by the two scenarios in terms of content distribution mainly fall into three categories:

• The user experience problem is mainly reflected in the complex environment of domestic operators and poor access effects across operators; Wide user distribution, high latency and slow speed of cross regional access; Poor transmission performance, slow speed, and severe congestion in weak network environments;
• The performance and cost issues of the source station are mainly reflected in the following aspects: the poor elastic expansion ability of the source station in a centralized deployment, unable to withstand large-scale burst traffic, and the high bandwidth cost of the source station in a centralized deployment;
• Business security issues are mainly reflected in: constantly facing security threats such as DNS hijacking (such as being hijacked for page content, being embedded in page advertisements, etc.), resource piracy (the source site faces losses in bandwidth costs), and so on.

2、 CDN content distribution network architecture

Currently, CDN has become a very important technical means to alleviate network congestion, improve business response speed, and improve user experience. It is also an indispensable and important component of Internet infrastructure. Below, we will illustrate the components of a typical CDN content distribution network by integrating the content distribution network architecture of Fireumbrella Cloud with CDN:

• Dispatching system

The user’s terminal will initiate various application accesses, such as two application scenarios: on-demand and download. The user’s access request will first be resolved through domain name DNS, and then the user’s domain name DNS resolution request will be processed through the Fireumbrella CDN scheduling system;

• Link Quality System

The role of the link quality detection system is to monitor the real-time load and health status of each node and link in the cache system in real time, and then feed back to the scheduling system. The scheduling system will analyze the user’s operator and region ownership based on the IP address information carried in the user request, and then allocate a good access node for the user based on the link quality information;

• Cache system

After receiving the IP address of the good access node, the user will access the cache node. If the node has already cached the resources requested by the user, it will directly return the resources to the user; If the L1 node fails to hit, it will be traced back to the L2 node. If the L2 node also does not have cache resources, it will return to the source station to pull this resource; After a resource is pulled, it will be cached in the cache system. This way, if subsequent users access the same resource, they can directly pull the resource from the cache system to avoid repeating the process of returning to the source; This deployment architecture of hierarchical caching can achieve the effects of improving content distribution efficiency, reducing back-to-source bandwidth, and improving the user experience.

• Support service system

Fire Umbrella Cloud CDN also provides a support service system in addition to content distribution. It has a resource monitoring capability that can monitor the status of customer business operations on the cache system, such as monitoring common indicators such as QPS, bandwidth, and http status codes for CDN accelerated domain names. The second is data analysis, where users can analyze the TOP URL, PV, UV, and other data of CDN accelerated domain names. The third is the configuration management system, through which users can issue cache rules such as cache file type, cache time, and de parameter cache to improve the operational efficiency of the cache system.

3、 How can CDN solve the problems faced by Internet content distribution?

Next, let’s look at how CDN solves the problem from three aspects: user experience, origin site performance and cost, and business security.

（1） User experience issues

Firstly, CDN can provide users with many nearby access nodes. Fireumbrella Cloud CDN integrates more than 20 mainstream CDN manufacturers in the world, basically covering major domestic and overseas operators. Users can freely choose from up to 14000 high-quality nodes, conduct intelligent scheduling across the network, and achieve high-quality services. Relying on rich resource node coverage, Fireumbrella Cloud CDN can enable users to access nodes with the same operator nearby, solving user experience issues such as interoperability across operators and large long-distance access latency.

Secondly, the second point to improve the user experience is the need to provide accurate scheduling capabilities. Here, Fireumbrella Cloud CDN provides five key optimization measures:

• Continuously updated accurate IP database: The role of the IP database is to determine the user’s region and operator affiliation when the user’s DNS resolution request is forwarded to the CDN scheduling system, thereby allocating the user’s nearest access to the same operator CDN node. To ensure that the data in the IP database is in a new state, the database is constantly updated.
• Provide HTTPDNS service (requires client compatibility): Using HTTPDNS technology allows user terminals to bypass the operator’s local DNS and directly use the HTTP protocol to access the scheduling system, requesting optimal access nodes for the domain name they need to access, which can avoid the business security issues caused by DNS hijacking.
• Node data analysis: The CDN scheduling system analyzes the health status of all nodes and links in the entire cache system in real time through the link state system, selecting optimal access nodes for users, to avoid affecting the user’s access experience due to poor access node quality.
• Content based scheduling to improve access hit rates: In two application scenarios, large file downloads and video on demand, 302 scheduling is often used as a content based scheduling technique. 302 scheduling is a central scheduling scheme. When a user requests a resource, after completing DNS resolution of the domain name, the user’s request will first access the central scheduling system, The central scheduling system will then parse the content that the user requests to access, and then assign a good access node to the user through 302 redirection.
• The company has 100+monitoring nodes, and cooperates with professional network monitoring companies, channels, and customer monitoring systems to grasp the status of CDN nodes, operator backbone networks, customer origin sites, HTTPS certificates, DNS resolution, and other conditions in real time. Only by collecting sufficient data information can a more scientific basis for scheduling decisions be achieved.

By adopting the above five optimization measures, users can be provided with more accurate scheduling capabilities, allowing them to access good CDN nodes.

A simple comparison of the three common scheduling schemes for CDN is shown in the following figure:

Here is a brief explanation of the advantages and disadvantages of these three common CDN scheduling methods:

① DNS scheduling is a commonly used and general scheduling scheme, with the disadvantage that there is a risk of DNS hijacking and the degree of scheduling may be poor;

② 302 scheduling is very suitable for use in two application scenarios: large file download and video on demand. The advantage is that it can improve the degree of scheduling, but the disadvantage is that it will increase the delay of the first packet (in large file download and video on demand scenarios, it is not sensitive to the delay of the first packet, and requires higher scheduling degree)

③ The advantage of HTTP DNS scheduling is that it has high security (which can avoid the risk of DNS hijacking) and scheduling, but it has a significant disadvantage that it requires client support (such as embedding an SDK on a mobile app), which is less versatile.

Thirdly, the third optimization point for improving the user experience is intelligent optimization of the transmission link

• Protocol optimization: By adjusting congestion algorithms, packet loss detection algorithms, and other optimization measures, the performance of the TCP protocol stack has been significantly improved, especially in weak network environments; The following two bandwidth utilization comparison charts show the standard TCP transmission protocol on the left. After establishing a TCP connection, it takes a relatively longer time to improve bandwidth utilization. Once packet loss occurs, bandwidth utilization will rapidly decline, and it needs to undergo a slow ramp stage to gradually improve bandwidth utilization, especially in weak network environments where standard TCP protocols are used, The decline in bandwidth utilization can be very serious. On the right is the optimized TCP protocol for Fire Umbrella Cloud. Through technical optimization measures such as rapid startup, active congestion detection, and rapid packet loss recovery, it can provide higher bandwidth utilization, and even in relatively poor network conditions, it can have relatively good optimization effects.

• Connection acceleration: Fast retry is achieved by modifying the Handshake Timer of the protocol stack to compensate for the retry timeout caused by packet loss;
• Persistent connection: Establish a TCP connection between CDN nodes in advance and keep it open for a long time, so that when new connections come over, they can be directly reused, reducing TCP handshake latency.

（2） Source station performance and cost pressure issues

Helping customers reduce the performance and cost pressures of the origin site by improving the efficiency of the cache system mainly includes six technical measures:

• Distributed deployment: Ensure that users can access nodes with the same operator nearby, thereby improving access speed and solving problems such as slow access across operators and high long-distance back-to-source latency;
• Multilevel cache architecture: Each level of the multi-level cache architecture can be used to cache resources requested by users. After multi-level backsource convergence, backsource traffic can be greatly reduced;
• Content prefetching: Prefetching the content of the source station to the CDN node in advance through content prefetching technology, reducing the amount of back-to-source and improving the cache hit rate;
• Merge back-to-source requests: Configure the merging of back-to-source requests for several domain names with a certain domain name (under the same service) to reduce repeated back-to-source traffic;
• Deparameter caching: When setting a hashkey (cache key), configure to remove the parameters following the resource URL and use a URI without parameters as the cache key to improve the cache hit rate;
• Flow control: Provide refined flow control capabilities, support single request flow control and network wide flow control, support edge flow control and backflow flow control, and effectively reduce customer costs.

The serial numbers marked in the following figure represent the role of the above six technical measures on the CDN cache system:

Taken together, these technical measures can help customers effectively reduce back-to-source bandwidth and improve cache hit rates, thereby reducing bandwidth pressure and reducing bandwidth costs at the origin.

（3） Business security risk issues

Fire Umbrella Cloud CDN helps users avoid business security risks through reliable security measures.

1. Anti-theft chain: You can choose to use universal authentication methods such as referer, UA, URL, and IP, or you can use EdgeScript to customize authentication rules, support edge authentication and central authentication, and strictly verify user resource access requests to prevent the source site resources from being embezzled;

2. DNS anti hijacking: HTTPDNS technology uses the HTTP protocol to access the server side of Fireumbrella Cloud to obtain domain name resolution results, which can bypass the operator’s Local DNS to avoid domain name DNS hijacking;

3. HTTPS transmission encryption: Support the use of TLS protocol to encrypt HTTP protocol content, prevent clear text data from being exposed on the Internet, and can set functions such as TLSv1.3 and HSTS.

CDN Accelerated Practice of Fire Umbrella Cloud CDN in Large File Download and Video on Demand

4、 What core capabilities does Fireumbrella Cloud CDN provide in two major application scenarios: large file download and video on demand

Rich resource nodes: Same operator CDN nodes that can provide users with nearby access, solving the problems of high latency and slow speed caused by long-distance access and cross operator access

Accurate scheduling system: It can obtain the health status of CDN nodes in real time, and allocate appropriate access nodes based on the user’s location and operator to achieve good access effects

Intelligent transmission link: Reduce overall delay and improve transmission speed through measures such as protocol optimization and connection optimization, especially in weak network environments

Efficient caching strategy: An efficient caching strategy can lead to a higher user hit rate (hitting cache resources on nearby nodes), thereby providing higher access speed

Reliable security protection: Reliable security measures can strictly verify and encrypt user resource access requests, protecting user resources and business security