WO2005022825A1

WO2005022825A1 - A system of layered content delivery network and a method of layered delivering content

Info

Publication number: WO2005022825A1
Application number: PCT/CN2004/000514
Authority: WO
Inventors: Anyuan Xiong; Daiyi He
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2003-08-29
Filing date: 2004-05-21
Publication date: 2005-03-10
Also published as: CN1592248A; CN100349431C; RU2321956C2; HK1074937A1; RU2006110034A

Abstract

The invention discloses a layered content delivery network, the network comprises content provider layer, content delivery layer including at least one delivery sublayer and edge server layer; content provider layer comprises at least one ICP node, the content provider layer is connected straightly to the first delivery sublayer in the content delivery layer; each delivery sublayer in the content delivery layer comprises at least one delivery node, the last delivery sublayer in the content delivery layer is connected straightly to edge server layer; edge server layer comprises at least one edge server region; the content provided by content provider layer is delivered down to edge server layer step by step through the delivery node in the first delivery sublayer connected straightly to content provider layer. At the same time, the invention also discloses a method of layered delivering content. Application of the invention makes CDN makeup network method more flexible, optimizes traditional CDN attemper method, advances CDN server quality, and can realize distributed server and realize region equipoise of server load, and can advance the visit shoot rate.

Description

TECHNICAL FIELD The present invention relates to a hierarchical content distribution network and a method for hierarchical content distribution.

The present invention relates to a content distribution network technology, and in particular, to a layered content distribution network and a method for layered content distribution. Background of the invention

With the rapid growth of Internet users and the increasing demand from users for improving the quality of Internet services, the traditional model of directly providing information services by Internet Content Providers (ICPs) is facing serious challenges. ICP application sites generally establish physical sites in one or a few places to provide information services to Internet users nationwide and even globally. The Internet's own structural characteristics and network bandwidth bottlenecks make people visit global sites with long delays and low efficiency.

Information distribution is an effective way to solve these problems, that is, by establishing a new layer of network on the basis of the existing Internet-Content Delivery Network (CDN, Content Delivery Network), which is specifically used to efficiently deliver rich multimedia through the Internet content. CDN greatly improves the response speed of the network by realizing the user's near access to the website and the intelligent distribution of network traffic, and technically comprehensively solves the problems caused by the user due to small network bandwidth, large user access, and uneven distribution of outlets. Issues affecting access performance. The technical principle of the CDN is to push the content of the website from the backbone of the network to the "edge" of the network closest to the user in the existing Internet network, in order to reduce the factors affecting the effect such as transmission delay and network jitter, thereby providing orderly And high-quality services allow users to get the information they need from the place closest to the user as quickly as possible. Because this technology has greatly eased the congestion of the Internet, the website has the ability to provide more content services with huge data traffic such as video programs, song on demand, etc. At the same time, the reliability of various services such as online transactions and online banking is also available. Get effective Protection. In short, the CDN can publish the information of the ICP site to the edge devices of the network in advance, so that users can get a quick response on the nearest edge device when visiting the site, thereby improving the quality of service to users and reducing the center. Node load.

The current CDN networking structure is relatively simple. Whether it is enterprise-wide, metropolitan-wide, or nationwide, a single-level structure is used. The network structure is shown in Figure 1. The main components of a CDN are devices such as a global load balancer, a central server (CS), and an edge server (ES). The network architecture can be divided into two parts: the center and the edge. The central part is generally composed of devices such as the global load balancer and the central server. The edge part is generally composed of many edge servers as the edge device part that provides users with access information. The information of the ICP site is pre-published to each edge server through the central server, and each user's access request to the ICP site is scheduled by the global load balancer. From the CDN network structure shown in Figure 1, it can be seen that although the CDN's edge servers are distributed at the edge of the network, the global load balancer as a scheduler is still concentrated in the center of the network.

In the above-mentioned CDN network structure, the general user access scheduling process is shown in Figure 2, and includes the following steps:

Step 201: The user accesses a WEB website through a browser, and clicks content in the website.

Step 202: The browser of the user terminal requests a local domain name server (DNS, Domain

Name Server) to resolve the domain name of the website.

Step 203: The local DNS requests resolution of the authorized DNS of the website.

Step 204: The authorized DNS of the website notifies the local DNS, and the authorized DNS of the subdomain name of the website domain name is a global load balancer, and returns the IP address of the global load balancer.

Step 205: The local DNS sends a resolution request to the global load balancer.

Step 206: The global load balancer selects an optimal ES based on a certain preset policy, usually in a round-robin manner, and returns the IP address of the ES to the local DNS. Step 207: The local DNS returns the IP address of the ES to the browser of the user terminal.

Step 208: The user accesses the ES according to the returned IP address, and sends a request to the ES. Step 209: If the content required by the user does not exist on the ES, the ES obtains the content required by the user from the central server, and then execute step 210; otherwise, directly perform step 210.

Step 210: The ES provides the required content according to the request of the user.

Although the CDN adopting the above scheme improves the user's access quality to some extent, there are still some problems, which make the CDN far from being perfect in terms of service quality and network use efficiency. For example: ① It is difficult to perform large-scale CDN Networking; ② Because the unified centralized scheduling method is adopted, no matter how wide the coverage area is, the global load balancer as the scheduler is centralized, resulting in heavy load on the scheduler, slow response, and concurrent load on the scheduler The ability requirements are also very high; ③ Cannot detect the user's proximity based on the user's IP address and accurately schedule the nearest location; ④ In content release management, only pull-down (PULL) is supported, PUSH is not supported, and the ability to actively push is weak; ⑤ When the edge server is selected, the content is not detected, resulting in a low access hit rate, and the network busy status is not detected, and a network with small traffic cannot be selected to provide services.

At present, there is another solution using a Layer 4 (L4) switch. In this solution, the L4 switch has the function of the global load balancer in the above solution. The general user access scheduling process for this solution is shown in Figure 3 and includes the following steps:

Step 301: A user accesses a WEB website through a browser, and clicks content in the website. Step 302: The browser of the user terminal requests the local DNS to resolve the domain name of the website.

Step 303: The local DNS forwards the resolution request to the L4 switch.

Step 304: The L4 switch requests resolution of the authorized DNS request from the website.

Step 305: The authorized DNS of the website resolves the IP addresses of multiple ESs according to the request. These IP addresses are returned to the L4 switch.

Step 306: The L4 switch selects an optimal ES among multiple ESs according to a certain preset policy, such as a round-robin manner, and returns the IP address of the ES to the local DNS.

Step 307: The local DNS returns the IP address of the ES to the browser.

Step 308: The user accesses the ES according to the IP address returned by the local DNS and sends a request to the ES.

Step 309: If the content required by the user does not exist on the ES, the ES obtains the content required by the user from the central server, and then execute step 310; otherwise, directly perform step 310.

Step 310: The ES provides the required content according to the request of the user.

There are still some problems with the CDN adopting the above solution, for example: all accesses are concentrated to the L4 switch, and its load is too heavy; for user requests, the L4 switch can only process the information of the fourth layer of the network protocol, but not the seventh Information, that is, the content cannot be detected.

From the above two solutions, it can be seen that with the increase of users, the currently widely used single-level CDNs are no longer able to adapt to large-scale and high-density application environments, and cannot achieve distributed services and achieve geographical balance of business loads. Summary of the invention

In view of this, the main object of the present invention is to provide a layered content distribution network and a method for content layered distribution, so that the CDN networking method is more flexible, the traditional CDN scheduling method is optimized, and the CDN service quality is improved, It can achieve distributed services and regional balance of business load, and can improve the hit rate of access.

To achieve the above object, the technical solution of the present invention is implemented as follows:

A hierarchical content distribution network includes a content distribution layer, a content distribution layer including at least one level distribution sublayer, and an edge service layer; the content distribution layer includes at least one ICP node, the content distribution layer and content The first-level distribution sub-layer in the distribution layer is directly connected; Each level distribution sublayer of the content distribution layer is composed of at least one distribution node, and the lowest level distribution sublayer in the content distribution layer is directly connected to the edge service layer; the edge service layer is composed of at least one edge service area; The content that the content publishing layer needs to publish is distributed to the edge service layer level by level through the distribution nodes in the first-level distribution sublayer directly connected to it.

Each of the ICP nodes includes at least one ICP operation terminal and at least one ICP. Each said distribution node includes at least one media manager (MM) and one central server (CS).

Each edge service area includes at least one CS, one MM, one media request scheduler (MRB), and at least one edge server (ES).

The MRBs in all edge service areas are distributed in the local center of each edge service area.

The ICP node in the content publishing layer is a node above the distribution node in the first-level distribution sub-layer directly connected to it, and the edge service area in the edge service layer is the lowest-level distribution sub-layer directly connected to it The next-level node of the distribution node in.

A method for hierarchical distribution of a content distribution network includes the following steps:

a. The ICP node distributes the content to be published to at least one distribution node in a content distribution layer directly connected to the ICP node;

b. In the content distribution layer, the distribution node that currently receives the content distribution command determines whether it is the distribution node in the lowest-level distribution sub-layer, and if so, performs step c; if not, then directly connects to it Read the required publishing content in the upper node, store it on the CS in this node, and distribute the content to one or more lower-level distribution nodes directly connected to it, and return to step b;

c. According to the received content distribution command, the current distribution node receives the content to be published from the node directly above it, stores it on the CS in this node, and distributes the content to one or one directly connected to it. Above edge service area. The method further includes: the user's access to the ICP is scheduled by the MRB in the edge service area where the user is located, and the CS or ES in the edge service area where the user is located provides the service required by the user.

The scheduling includes at least analyzing the domain name of the ICP accessed by the user, detecting the ES in the edge service area, or selecting the best ES.

The detection of the ES includes at least detecting the proximity of the ES to the user, the presence or absence of content required by the user, or the network busy status.

The selection is to determine the best ES based on the proximity of the ES to the user, the presence or absence of user-required content, or the availability of the network.

Compared with the prior art, networks of various sizes can be constructed by applying the present invention, including metropolitan area networks, provincial networks, or national networks. It also supports step-by-step capacity expansion from the metropolitan area to the whole province, and even nationwide, making the CDN networking more flexible. The invention optimizes the traditional CDN scheduling method, adopts distributed scheduling, reduces the service scope of the scheduler, reduces the load, shortens the response time, and can further improve the service quality of the CDN network. Since the present invention adopts a hierarchical and hierarchical content distribution management, thereby ensuring that ICP requires only a little access, it can realize the entire network release and the entire network services, thereby achieving distributed services and achieving regional balance of business load. At the same time, the present invention can perform scheduling based on the busy status of the edge server and the proximity of the user, select a relatively free edge server to provide services, and improve service quality; it can also perform scheduling based on content to improve the hit rate of access. Brief description of the drawings

FIG. 1 is a schematic diagram of a network structure of a prior art solution;

2 is a schematic diagram of a user access scheduling process in the first solution of the prior art;

FIG. 3 is a schematic diagram of a user access process in the second solution of the prior art;

4 is a schematic diagram of a network structure according to an embodiment of the present invention; 5 is a schematic diagram of a content publishing process according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of a user access scheduling process according to an embodiment of the present invention. Mode of Carrying Out the Invention

In order to make the objectives, technical solutions, and advantages of the present invention clearer and clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

The main idea of the present invention is: Use distributed scheduling and hierarchical distribution to form a network. The scheduler is not concentrated in the center of the network, but is deployed in the local center of each service area and is responsible for access scheduling in this area. Contents are distributed and managed in a hierarchical manner, ensuring that ICP requires only one point of access to publish across the entire network and provide services across the entire network.

The hierarchical CDN according to the present invention is composed of a content publishing layer, a content distribution layer including at least one level distribution sublayer, and an edge service layer. Taking the network structure in which the content distribution layer is composed of two level distribution sublayers as an example, The network structure is shown in FIG. 4: The main components of the CDN are ICP, ICP operation terminal, central server (CS), media manager (MM), media request scheduler (MRB), ES and other equipment. Among them, the MM is responsible for processing signaling, and its main function is to accept a release command on content release issued by the ICP operator, and issue a content distribution command to the ES, the next-level MM, or the CS at all levels according to the release command, and receive the ES And feedback information from MMs or CSs at all levels; CSs at all levels are responsible for the content storage, distribution, and service provision of the center; MRBs in all edge service areas in the edge service layer are distributed in parts of each edge service area The center is mainly responsible for access scheduling within the local network to achieve load balancing. The solid line in FIG. 4 indicates the process of content distribution and distribution; the dashed line indicates the content distribution command.

The various components in the above CDN can be combined into: an ICP node, a distribution node, and an edge service area. Among them, each ICP node includes at least one ICP operator, one or more ICPs; each distribution node includes at least one MM, one CS; each edge service area includes at least one CS, one MM, one MRB, and one or Multiple ES; The content publishing layer is composed of one or more ICP nodes, the content distribution layer is composed of one or more levels of distribution sublayers, each level of distribution sublayers is composed of at least one distribution node, and the edge service layer is composed of one or more edge service areas . 'The content distribution layer is directly connected to the content distribution layer, the content distribution layer is directly connected to the edge service layer, and the content to be distributed by the content distribution layer is distributed to the edge service layer through the content distribution layer. Distribution nodes are divided into different levels: The distribution nodes that are directly connected to the content distribution layer are regarded as the first-level distribution nodes. The collection of all first-level distribution nodes is called the first-level distribution sublayer, and the ICP nodes in the content distribution layer are directly connected to it. The upper-level node of the first-level distribution node; the distribution node that has the same role as the first-level distribution node directly is called the second-level distribution node, and the set of all second-level distribution nodes is called the second-level distribution sub-layer, and the first-level distribution The node is the upper-level node of the second-level distribution node directly connected to it; and so on, the edge service area of the edge service layer is directly connected to the lowest-level distribution node, and the collection of all the lowest-level distribution nodes is called the lowest The first-level distribution sub-layer, and the lowest-level distribution node is the upper-level node of the edge service area directly connected to it. There is at least one primary distribution node in the content distribution layer.

The content required to be published by the content publishing layer is first distributed to one or more first-level distribution nodes that are directly connected to the first-level distribution node. Multiple second-level distribution nodes, and then the second-level distribution node distributes the received published content to one or more third-level distribution nodes that are directly connected to the second-level distribution node; and so on, until the required The published content is distributed down to the edge service layer level by level, so as to realize the hierarchical distribution of content.

Based on the CDN network structure shown in Figure 4, the general content publishing process is shown in Figure 5, and includes the following steps:

Step 501: The ICP operation end of an ICP node in the content release layer issues a content release command to the MM in one or more first-level distribution nodes directly connected to it. Step 502: The MM in the first-level distribution node issues a content distribution command to the CS in the distribution node where the MM is based on the received content distribution command.

Step 503: The CS in the first-level distribution node reads the content that the ICP needs to publish from the ICP content storage device of the ICP node according to the received content release command, and stores the content to the first-level distribution node. CS on.

'Step 504, the CS in the first-level distribution node notifies the MM in the distribution node where it is located, and the content that the ICP node needs to publish has been stored on the CS in the first-level distribution node.

Step 505: The MM in the first-level distribution node issues a content distribution command to the MM in the second-level distribution node directly connected to the node.

Step 506: The MM in the secondary distribution node sends a content publishing command to the CS in the node where the MM is located according to the received content publishing command.

Step 507: According to the received content distribution command, the CS in the second-level distribution node reads the content to be published from the upper-level CS, that is, the CS in the first-level distribution node, and stores the content in the second-level distribution node. CS.

Step 508: The CS in the secondary distribution node notifies the MM in the node where the CS is located, and the content to be published has been stored in the CS in the secondary distribution node.

Step 509: The MM in the secondary distribution node sends a content release command to the MM in the edge service area directly connected to the node.

Step 510: The MM in the edge service area issues a content release command to the CS in the edge service area where the MM is located according to the received content release command.

Step 511: According to the received content distribution command, the CS in the above-mentioned edge service area reads the content to be published from the upper-level CS, that is, the CS in the second-level distribution node directly connected to the CS, and stores the content in the local CS, that is, Above the CS in the edge service area.

Step 512: The CS in the edge service area notifies the MM in the node in which it is located that the content to be published has been stored on the CS in the edge service area. Step 513: The MM in the edge service area issues a content distribution command to the ES in the edge service area where the MM is located.

Step 514: The ES in the edge service area reads the content to be published from the CS in the edge service area where the ES is located according to the received content distribution command, and stores the content on the ES.

After the above content publishing process, the content that ICP needs to publish is pushed to the edge of the network, that is, the ES in the edge network, through the above-mentioned hierarchical CDN. The above content publishing process applies not only to provincial CDNs, but also to metropolitan CDNs and national CDNs.

In the CDN network structure according to the present invention, a general user access scheduling process is shown in FIG. 6 and includes the following steps:

Step 601: The user accesses a WEB page and clicks on the content of the website.

Step 602: The browser of the user terminal requests the local DNS to resolve the domain name of the website by the user.

Step 603: The local DNS requests resolution of the authorized DNS of the website.

Step 604: The authorized DNS of the website notifies the local DNS, the authorized DNS of the corresponding content of the website is the MRB deployed locally, and returns the IP address of the MRB.

Step 605: The local DNS returns the MRB IP address to the browser.

Step 606: The browser of the user terminal accesses the MRB.

In step 607, the MRB detects all ESs in the area, selects the best ES based on the proximity and content, and returns the IP address of the ES to the browser.

Step 608: The user accesses the ES and makes a request to the ES.

Step 609: If the content required by the user does not exist on the ES, the ES obtains the content required by the user from the local CS, and then execute step 610; otherwise, directly perform step 610.

Step 610: The ES provides the required content according to the request of the user.

It can be seen from the above process that after the local user's access is resolved through DNS, its access is both scheduled by the local MRB and provided by the local ES, which reflects the business well. The idea of distributed implementation.

Because the present invention adopts distributed scheduling and hierarchical distribution for networking, compared with the traditional CDN, the present invention has the advantage of supporting hierarchical hierarchical networking, which can be used to build a small-scale CDN or a large-scale CDN. CDN; The content distribution management of the present invention adopts a hierarchical structure, so as to ensure that the ICP can be released across the entire network and serve the entire network with only a little access; Because the present invention divides the service area, each edge service area is distributed locally. Scheduling, thereby effectively reducing the load on the scheduler, and can also illustrate the status of the network. The present invention is also applicable to a network structure in which the content distribution layer is composed of one or more sub-layers.

The above description is only the preferred embodiments of the present invention, and is not intended to limit the protection scope of the present invention.

Claims

Claim

1. A hierarchical content distribution network, characterized in that the network is composed of a content distribution layer, a content distribution layer including at least one level distribution sublayer, and an edge service layer; the content distribution layer includes at least one ICP node, The content publishing layer is directly connected to a first-level distribution sublayer in the content distribution layer; each level of the content distribution layer is composed of at least one distribution node, and the lowest-level distribution sublayer and edge in the content distribution layer The service layer is directly connected; the edge service layer is composed of at least one edge service area; the content that the content publishing layer needs to publish is distributed downward to the edge service layer through the distribution nodes in the first-level distribution sublayer directly connected to it.

2. The content distribution network according to claim 1, wherein each of said ICP nodes includes at least one ICP operation terminal and at least one ICP.

3. The content distribution network according to claim 1, wherein each of said distribution nodes includes at least one media manager (MM) and one central server (CS).

4. The content distribution network according to claim 1, wherein each edge service area comprises at least one CS, one MM, one media request scheduler (MRB), and at least one edge server (ES).

5. The content distribution network according to claim 4, characterized in that: MRBs in all edge service areas are distributed and deployed at a local center of each edge service area.

6. The content distribution network according to claim 1, wherein: the ICP node in the content publishing layer is a node above the distribution node in the first-level distribution sub-layer directly connected to the edge service, The edge service area in the layer is the next-level node of the distribution node in the lowest-level distribution sub-layer directly connected to it.

7. A hierarchical distribution method for a content distribution network, characterized in that the method includes the following steps: a. The ICP node distributes the content to be published to at least one distribution node in a content distribution layer directly connected to the ICP node;

b. In the content distribution layer, the distribution node that currently receives the content distribution command determines whether it is the distribution node in the lowest-level distribution sub-layer, and if so, performs step c; if not, then directly connects to it Read the required publishing content in the upper node, store it on the CS in this node, and distribute the content to one or more lower distribution nodes directly connected to it, and return to step b;

c. According to the received content distribution command, the current distribution node receives the content to be published from the node directly above it, stores it on the CS in this node, and distributes the content to one or one directly connected to it. Above edge service area.

8. The method according to claim 7, further comprising: the user's access to the ICP is scheduled by the MRB in the edge service area where the user is located, and is performed by the user in the edge service area where the user is located. CS or ES provides the services required by the user.

9. The method according to claim 8, characterized in that: the scheduling comprises at least resolving the domain name of the ICP accessed by the user, detecting the ES in the edge service area or selecting the best ES.

10. The method according to claim 9, characterized in that: said detecting the ES comprises at least detecting the proximity of the ES to the user, the presence or absence of content required by the user, or the status of network busyness.

11. The method according to claim 9, wherein the selection is to determine the best ES based on the proximity of the ES to the user, the presence or absence of content required by the user, or the network busy status.