RU2454711C1 - Method of distributing load between content delivery network (cdn) servers - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: method of distributing load between content delivery network (CDN) servers where a service request is received from a user terminal on a server, the user address is determined, the CDN server address selected from a plurality of CDN server addresses is brought into conformity with the user address and the route for communication of the user with the corresponding CDN server is selected, wherein bringing into conformity is carried out using a data base of candidate routes formed on the server, wherein selection of the route is carried out based on a route metric selected from: delay, delay variation, load, packet loss percentage (or number), number of self-contained systems, Q-criterion, wherein the interval of permissible values for each route metric is further provided, and if the route metric is not in said interval, the respective route is excluded from the corresponding data base of candidate routes and, also, activity of all CDN servers is monitored in real-time using a means of detecting faults, connected to a server, wherein upon detection of fault conditions of the CDN server, information on status and address of such a CDN server is entered into the corresponding data base of the server and subsequently not considered during route selection.

EFFECT: high rate of delivering content to a user; high fault-tolerance of the CDN system.

2 cl, 1 dwg

Description

Technical field

The claimed invention relates to the field of CDN networks (content delivery networks), and more specifically, to a method of load balancing between servers in such networks.

State of the art

A conventional method for providing services to a user is known from the prior art, such as providing downloading content (e.g., site content) to a terminal (desktop computer, mobile phone, etc.) of a user, in which said content is placed on a server on a network (e.g., network) Internet), and users who want to use the services connect to this server and receive the desired content via communication networks.

The disadvantages of the traditional method are the long waiting time and low download speed of the content if a large number of users connect to the server at the same time. This circumstance is due to the limited bandwidth of the channel through which the server is connected to the network.

In addition, these drawbacks can manifest themselves if the server is geographically located far enough from the connected user. In this case, on the way from the server to the user, the data signal, according to the existing data transfer protocols, undergoes several routing steps, which, accordingly, increases the number of transit sections (hopes) and leads to a transmission delay.

In the rapidly growing market for the provision of multimedia services, in which the content is a large amount of data (video, music, images, etc.), the application of the traditional method becomes unacceptable for the requirements of the quality of user service.

To overcome these drawbacks of the traditional method, specialists developed CDNs, which are geographically distributed CDN servers, which can reduce the number of routing steps, as well as increase the speed of downloading content. A CDN typically consists of means for locating a user, means for selecting a nearby regional CDN server, CDN servers, central servers, and server failure detection means.

One embodiment of a CDN is as follows.

A user who wants to receive a service calls through a browser to a central server, which determines its location and redirects to the corresponding regional server via a communication protocol (for example, HTTP), where the user continues his work (receives the service). The user request is usually delivered to the central server via the network of the user provider, which has access to the CDN.

This option has drawbacks in the form of a slow speed of operation at the first user contact, the inability to redirect the user to another server in case of failure of the regional server.

Another embodiment of the CDN is the use of optimal route selection, which allows you to select the optimal regional server for a particular user and significantly increase the speed of service.

In accordance with this, the closest analogue of the claimed invention is a method of load balancing between CDN servers (WO 0184356 A2, 11/08/2001, G06F 17/00), in which a service request is received from a user on a central server, the user's location is determined by his the address, the CDN server address selected from the plurality of CDN server addresses is brought into correspondence with the user's location, and a route for communication between the user and the corresponding CDN server is set, and the CDN server address is selected based on Riki route selected from the group of averaged data transmission delay on the route, the average processing time information on a CDN-server path bandwidth, CDN-load server.

The disadvantages of this method is the insufficient assessment of the network situation when choosing a candidate route. In addition, this method does not provide mechanisms for flexible route changes in case of failure of a previously selected CDN server. In addition, in this method, the central server is a system vulnerability during a critical increase in load.

Thus, the object of the present invention is to overcome the above disadvantages of the prior art.

SUMMARY OF THE INVENTION

The above advantages are realized in that in a method of load balancing between servers of a content delivery network (CDN), in which a service request is received from at least one user terminal on at least one server, a user address is determined, brought into correspondence user address, CDN server address selected from a plurality of CDN server addresses. Next, a route is selected for user communication with the corresponding CDN server. The alignment is carried out by means of at least one database of candidate routes formed on at least one server. Moreover, the route selection is carried out on the basis of at least one route metric selected from the group: delay, delay variation, load, percentage (or number) of packet loss, number of hopes, number of autonomous systems, Q-criterion. In addition, an interval of acceptable values is provided for each route metric and, if the route metric is not included in the specified interval, the corresponding route is excluded from the corresponding database of candidate routes. In addition, they monitor the activity of all CDN servers in real time using a fault detection tool connected to at least one server. Moreover, in case of detection of the failure condition of the CDN server, information about the status and address of such a CDN server is entered into the corresponding database of at least one server and is then not taken into account when choosing a route.

In accordance with a particular case of execution, at least one server is additionally connected to appropriate means for caching content and corresponding means for collecting and analyzing statistical information.

The invention is illustrated in a schematic drawing, which shows the distribution of load between the servers of the content delivery network.

In accordance with a preferred embodiment of the invention, at least one server is provided in the CDN system, on which, using the means of collecting and storing information, a candidate route database (routing table) is formed for each of the system’s CDN servers, consisting of IP prefixes. Such a server may be, in particular, one or more CDN servers or one or more specialized central servers (backup servers). Further in the description, the term “server” means either a CDN or a specified central server. If several such servers are used in the system, their functionality is duplicated on each of these servers.

The specified database is formed, in particular, on the basis of data from open sources (GeoIP database, RIPE database, robtex.com website, lists of regional IP servers, etc.); information about the routes received from the operators where the CDN server is installed (both in static form and online using the BGP protocol).

In addition, each of the CDN servers can regularly, after a specified period of time (for example, 1 time per day), trace the route to a randomly selected IP address inside its IP prefix. At the same time, at least one of the following is collected: delay, variation of delay, load, percentage (or number) of packet loss, number of hopes, number of autonomous systems, Q-criterion (sum of the number of hopes and the number of autonomous systems). These data (metrics) are entered in the table of candidate routes of this server for further processing with the recording of this information in the database. Metrics can also be transferred to one or more CDN servers for processing and writing to the appropriate database.

In addition, for all CDN servers and for each of the candidate routes, the collected parameters can be checked for compliance with predetermined value ranges (thresholds). If this or that route is recognized as inappropriate according to the specified parameters, it is excluded from the database (s).

In addition, a blacklist of prefixes can be generated for each of the CDN servers, to which the server does not have the right to direct traffic. The list is formed on the basis of at least one of: analysis of log files (a large percentage of errors during file transfer, frequent cases of video buffering), based on information received from users or as a result of analysis carried out by system administrators of the CDN system. In addition, for all CDN servers, a check is made to see if the prefixes of candidate routes are blacklisted; if they do, such routes are eliminated.

In addition, for all CDN servers and for each of the candidate routes, the presence of information on the average connection speed for a particular IP prefix can be checked in the log files of the CDN server. If such information is statistically reliable, it is entered into the database of candidate routes.

In addition, all CDN system servers can be continuously monitored in the system to determine their activity through the fault detection tool. A CDN server is deprived of active status if this CDN server satisfies the failure condition: it is unavailable, or is overloaded, or too many percent of users receive refusals to display content, or too often the video is buffered. The specified information is sent to the server in the form of information about the status of the CDN server and its address, is entered into the database and then taken into account when choosing routes from the list of candidate routes.

In accordance with a preferred embodiment, upon receipt by the CDN system of a request from a user for a particular content, the selection of a CDN server for its delivery is carried out as follows: determine the user address and check whether the database has an IP prefix corresponding to the given user address. For this address, a route to the corresponding CDN server is determined. If there are several such prefixes, the prefix with the longest length is selected. If the same prefix has several routes, the prefix with the highest priority, depending on the metrics (listed below), is selected, unless the server status indicated by the corresponding row in the database is indicated as inactive.

In addition, when choosing a route, the route metrics discussed above are taken into account. Thus, the route is selected in accordance with the best metrics (for example, less latency, less server load, more bandwidth, etc.).

In an additional embodiment, the server (as well as the CDN server) can be additionally equipped with means for caching content that allow you to cache content that is most frequently accessed by users. Accounting of calls can be carried out by means of collecting and analyzing statistical information that can be located on servers (and CDN servers). The specified caching allows you to further increase the speed of content delivery to the user, reducing the number of hop-routes.

These tools can be a machine-readable medium with commands placed on it, under the control of which these servers collect and analyze statistical information.

From the foregoing, it will be clear to a specialist that the claimed invention provides a more complete assessment of the network situation when choosing a route for a particular user, which provides less delay when using the service.

In addition, real-time monitoring of the status of CDN servers allows you to build a route taking into account the characteristics of the content of the service, which is not provided by the current level of technology. In particular, the present invention provides for taking into account the quality of video display (including the buffering frequency) when determining the optimal route. Also, such monitoring allows to increase the fault tolerance of the CDN system (indicating the status of the CDN server), and increase its flexibility in general.

In addition, the present invention provides for taking into account the boundaries of the networks of Internet providers when determining the optimal route using the Q-criterion, which is the sum of the number of routers and autonomous systems on the route.

The use of the Q-criterion reflects the fact that the deterioration in the quality of communication on the intranet routers of Internet providers is less likely than the deterioration in the quality of internetwork routers installed at the borders of the networks of Internet providers, due to the greater load on the latter.

One skilled in the art will understand that the communication lines indicated in this application can be either wired or wireless communication lines, such as WiFi, CDMA, LTE, and the like.

In accordance with a preferred embodiment, a method is provided for load balancing between servers of a content delivery network (CDN), in which a service request is received from at least one user terminal on at least one server, a user address is determined, brought into correspondence the user address is the address of the CDN server selected from the plurality of addresses of the CDN servers, and a route is selected for communication of the user with the corresponding CDN server, the matching is carried out by means of, at least one database of candidate routes formed on at least one server, and the route is selected based on at least one route metric selected from the group: delay, delay variation, load, percentage ( or numbers) packet loss, the number of hopes, the number of autonomous systems, the Q-criterion, while additionally providing an interval of acceptable values for each route metric and, if the route metric is not included in the specified interval, exclude the corresponding a route from the corresponding database of candidate routes and, in addition, they monitor the activity of all CDN servers in real time using a failure detection tool connected to at least one server, and in case of detection of a failure condition of the CDN server , information about the status and address of such a CDN server is entered into the corresponding database of at least one server and is then not taken into account when choosing a route. In this case, the corresponding at least one database is located on the corresponding at least one server. It is obvious that the databases can duplicate each other, and each of the aforementioned CDN server and / or servers can be configured to exchange data with each other via respective communication lines.

One skilled in the art will understand that the communication lines indicated in this application can be either wired or wireless communication lines, such as WiFi, CDMA, LTE, and the like.

Claims (2)

1. A method of load balancing between servers of a content delivery network (CDN), in which a service request is received from at least one user terminal on at least one server, a user address is determined, the CDN address is mapped to the user address a server selected from a plurality of CDN server addresses and select a route for user communication with a corresponding CDN server, characterized in that the matching is carried out by means of at least one database of candidate routes, sf allocated on at least one server, and the route is selected based on at least one route metric selected from the group: delays, delay variations, load, percentage (or number) of packet loss, number of hopes, number of autonomous systems, the Q-criterion, in addition, provide an interval of acceptable values for each route metric and, if the route metric is not included in the specified interval, exclude the corresponding route from the corresponding database of candidate routes c and, moreover, they monitor the activity of all CDN servers in real time using a failure detection tool connected to at least one server, and in case of detection of a failure condition of the CDN server, information about the status and address of such CDN servers are entered into the corresponding database of at least one server and are then not taken into account when choosing a route. 2. The method according to claim 1, in which at least one server is additionally connected to appropriate means for caching content and corresponding means for collecting and analyzing statistical information.