TR202012584A2 - Server Assignment with Traffic Modeling in Content Distribution Networks (CDN) - Google Patents

Abstract

The invention is a performance enhancing system in Content Delivery Networks (CDN); - Data Collection Module (100) that collects data from CDN's PoPs, - Data Pre-Processing Module (200), which is used to make data available for use, - Traffic Modeling Module (300), which creates traffic modeling by performing self-similar analysis on CDNs, and - Characterized by the PoP Assignment List Module (400) that creates active PoP Assignment Lists (402) by processing the results of the Traffic Modeling Module (300) and data from other modules (100, 200).

Description

DESCRIPTION Server Assignment with Traffic Modeling in Content Delivery Networks (CDN) Technical Area Content Delivery Networks (CDN) It is a performance-enhancing method and related to the System. Prior Art Content requests from the end user in the current technique are different from the world. It reaches our DNS servers located in Predetermined DNS taking into account the user's location information are directed to the relevant POP according to the rules.

Fixed routing tables prepared according to these rules daily/monthly/yearly in internet traffic while forwarding your requests to POPs. It does not make statistical analysis of internet traffic changes and internet traffic. Like this Due to static POP lists, content transmission speed decreases and performance loss is being made. In addition, fixed PoP based on user location routing tables allow content over the more available PoP at the time. causes it to fail to transmit. So static POP routing tables It causes the user to access the content later. On the other hand; Static assignment list in PoPWar, browse internet bandwidth our company's internet band, as it directs without considering Width contracts are made according to the traffic where the internet is the most intense. This So, fixed POP assignment lists bandwidth POPs internet lease increases its costs.

Purpose of the Invention Content Delivery Networks (CDN); Picture, Video, Etc. contents, different geography to the Points of Presence (PoP) caches in the locations to deliver it quickly to users in different parts of the world by aims.

Ag data are Auto-Regressive according to their statistical self-similar characteristics. Active PoP assignment lists by giving outputs as input to Artificial Intelligence has been created. Self-similar analysis is generally accepted in the network field in the literature. Although it is a Phenomenon, self-Similar analysis in Content Delivery Networks (CDN) POP assignment list was not created by making traffic modeling. Contents Traffic modeling network operations and management in Distribution Networks (CDN) plays a key role in The purpose of the invention; Content Delivery Network (CDN) estimating the traffic volume by modeling the traffic and making the POP assignments is to ensure that it is done according to these data.

The purpose of the invention; fixed routing tables and end-user daily/monthly/yearly in internet traffic while forwarding your requests to POPs. making statistical analyzes of internet traffic is to provide.

Another purpose of the invention; millions of user requests, internet data routing according to active PoP assignment lists to be created using is to provide.

Another purpose of the invention; The traffic band on the Content Delivery Network (CDN) than POPs more available in terms of width, active `PoP routing lists' to reduce internet contract costs by routing.

The system developed to realize the mentioned purposes; Data Collection Module (100), Data Preprocessing Module (200), Traffic Modeling Module (300), and Includes PDF Assignment List Module (400).

Description of Figures Attached Figure - 1 of a CDN7 built on an existing internet network sematic view.

Figure-2 is the data collection and work flow diagram of the main module.

Figure-3 is the detailed work flow diagram of the inventive system.

The main elements expressed in the figures are listed below as numbers and names. given. (100) Data Acquisition Module (101) Bandwidth raw data (102) POP Raw Data (103) DNS Raw Data (104) POP measurements (200) Data Preprocessing Module (201) Bandwidth Data (202) POP Data (203) DNS Data (300) Traffic Modeling Module (301) Bandwidth Data (302) Self-Similarity Analysis (303) Training Data (304) Test Data (305) WINDING (306) Neural Network 1 (307) Model Parameters (308] Top Model (309] Conclusion (401) Neural Network 2 (402) POP Assignment List Detailed Description of the Invention The invention is in Content Delivery Networks (CDN) It is a performance enhancing product.

CDNts can be image, Video, Etc. content such as in different geographical locations. by caching in Points of Presence (POP) It aims to deliver it to users in different parts of the world quickly.

CDNs are more than many POPs built on the existing global internet network. is formed. POPs play an important role in content delivery: Well-structured With POPs, content delivery times are reduced and end users can access content more quickly. is reaching.

Content requests from end-users located in various parts of the world It reaches our DNS servers with the Anycast method. Anycast method same IP (Internet Protocol) address is valid for many servers; thus the last DNS query of the user to the nearest DNS server thanks to Anycast. is directed. End-user request after DNS server according to the predetermined DNS rules, taking into account the location information. It is directed to POP.

In systems in the current technique, depending on the location of the end user In which PoP to be routed is determined and unchanging POP assignment is done according to the order. End-user request on existing CDN systems Different weights that are not updated and assigned to different PoPs included in the CDNi They are guided by an algorithm.

Millions of user requests with this nation, using internet data It is directed according to the active PoP assignment lists to be created. traffic on CDN According to activePoP routing lists to more bandwidth-friendly PoPs Internet contract costs are reduced by directing

Ag data are Auto-Regressive according to their statistical self-similar characteristics. Active POP assignment lists by giving outputs as input to Artificial Intelligence has been created. Self-Similar analysis is generally accepted in the network field in the literature. Although it is a phenomenon, traffic modeling by self-similar analysis in CDNWer By doing this, the POP assignment list is not created. CDN traffic modeling, ag plays a key role in its operations and management. CDN With Invention It can predict traffic volume by modeling traffic and POP assignments are based on this data. is made accordingly.

A CDN built on top of the existing internet network is diagrammed in Figure 1.

CDN PoPs located in clusters of different Internet Service Providers (ISPs), It constitutes the Content Distribution Network. The subject of the invention is one of the building blocks of the System. The Data Acquisition Module (100) is located in each PoP and sends data to the Main Module. provides the flow.

There are four main modules in the inventive system: From CDN7's PoPis Data Collection Module (100), which collects incoming data, It transmits it to the Data Pre-Processing Module (200) for processing. Then Data Pre- The outputs of the Processing Module (200) are transmitted to the Traffic Modeling Module (300). Active POP Assignment by forwarding the incoming data to the POP Assignment List Module (400] Lists (402] are created. Finally, the POP Assignment List for each PoP (402) sent to its own PoP to be used for end-user routing The algorithm will come after processing the past time zone data for a certain period of time. transmits POP lists to POPs to be used in the time zone. Available The current data collected in the time period in which the POP lists are valid, creating input to the algorithm again and again at another time to come. It provides data entry by creating POP assignment lists (402) in the slice.

The Data Acquisition Module (100) transmits data to the next modules at certain intervals. PoP Assignment Lists (402) that will be sent and valid for a certain time period will be done. PoP Assignment Lists (402) at specific time intervals dynamic POP, which is the inverse of the fixed P0P Assignment Lists (402) Assignment Lists are made available on the CDN.

Bandwidth Raw Data from CDN POPs (101), PoP Raw Data (102), DNS Raw Data (103), RAM, CPU, Disk I/O Measurements collected from POPs (104); It is collected in the Data Collection Module (100). Collected raw data Collected PoP Measurements ( is being processed. Processed Bandwidth Data (201) to the Traffic Modeling Module ( if POP Assignment It is transmitted as input to the List Module (400).

In the Traffic Modeling Module (300), from the Data Preprocessing Module (200) The incoming Processed Bandwidth Data (201) is divided into time segments. To be used as Bandwidth Data (301) in the Modeling Module (300) is being printed. Partial Bandwidth Data (301), for statistical analysis Self- It is subjected to similarity analysis (302). After the aforementioned analysis, fragmentary Bandwidth Data is split appropriately into Training Data (303) and Test Data. Data (304) is obtained. Training Data (303); Seasonal Auto-Regressive (306) constructed and analyzed according to. At the end of the mentioned analyzes, Education Datasf (303) has the most suitable Model Parameters (307). of models After the parameters are determined, the Best Model is determined by using the Test Data (304). (308) is found by performing error analysis. Optimal model (308) error analysis after the result; Results of the Traffic Modeling Module (300) It is sent as an input to the POP Assignment List Module (400].

In the POP Assignment List Module (400); From Data Collection Module (100) incoming POP measurements (104), POP Data from the Data Preprocessing Module (200) Result ( to Neural Network Z to generate (401]. At the end of the aforementioned process, a specific time for each POP The P0P Assignment List (402) with the interval criteria is created. To POPs in CDNi, POP Kinetics by uploading Assignment Lists and updating them at specific time intervals The POP Assignment List (402) is now available.

The processes carried out by the inventive system are as follows; - Data Acquisition Module (100) to Data Pre-Processing Module (201) Transmission of Bandwidth Raw Data (101), - Data Acquisition Module (100) to Data Pre-Processing Module (201) Transmission of POP Raw Data (101), - Data Acquisition Module (100) to Data Pre-processing Module (201) Transmission of DNS Raw Data (101), - From Data Acquisition Module ( POP Measurements (transmission, - Ready-to-use Bandwidth in the Data Preprocessing Module (200) Transmission of Datas1 (201) to Traffic Modeling Module (300), - POP Data ready to use in Data Preprocessing Module (200) (transmission, - Ready-to-use DNS Data in Data Preprocessing Module (200) (transmission, - By dividing the Band Gemsligi Data (301) into appropriate time intervals Transmission for Self-Similarity Analysis (302), - Training Data of some of the Self-Similar Analysis (302) data (303) stored as, - Test Data of a part of the Self-Similar Analysis (303) data (302) stored as, - The method of “Training Data” and Artificial Neural Network (306) constructed and modeled, - Model of YELLOW MA (305) method and Artificial Neural Network 1 (306) results Determining its parameters (307), - The error margins of the traffic model with Model Parameters (307) to be forwarded to the Best Model (308) for calculation, - Bandwidth Test Data°n1n (304) for calculating margins of error forwarding to the best Model (308), - With the most pleasant margin of error, so that the most valid Traffic Modeling Result (To Artificial Neural Network 2” (401) transmission, - POP assignment lists of the results of Neural Network 2 (401) send it to the POP Assignment List (402) to create is performing.

The bandwidth raw data (101) included in the invention are the updated data from the PoPs. Indicates Bandwidth data.

POP Raw Data (102) represents current POP Data from POPs is doing.

DNS Raw Data (103), Expressing up-to-date DNS Data from POPs is doing.

CPU, RAM, Disk Input/Output data from PoPs POP Measurements It is collected at the point (104).

Bandwidth Data (201) Processed Bandwidth Raw Data (101) represents the state. Bandwidth Data (201), Traffic Modeling The module (300) is the input.

POP Raw data ( input generates POP Data (202).

DNS Raw data (to be input generates DNS Data (203).

Bandwidth Data (301) from Data Pre-Processing Module (200) divided into appropriate time intervals.

Band divided into time segments by Self-Similarity Analysis (302) method Statistical analysis of the Width Data (301) is performed.

Some of the available Bandwidth data (301) is a Training Data (303) part of it is reserved as a Test Data (304).

Seasonal Auto-Regressive using Bandwidth Training Data (303) is being done. Also, using the mentioned Bandwidth Training Data (303) Traffic modeling is also done by building Artificial Neural Network 1 (306). made By using models, SARIMA (305) and Artificial Nerve Ag1 have the least error. model parameters (307) are determined.

Error using Model Parameters and Bandwidth Test Data By analyzing it, is SARIMA (305) or the most method in traffic modeling? It is determined whether the Artificial Neural Network-1 (306) will be used. The mentioned error done by analysis. As a result, the best model (308) is determined.

POP Assignment List by calculating the best traffic modeling results (309) It is sent as input to the module (400).

From the Data Acquisition Module* ( and From the Data Pre-Processing Module ( with Using the results from the Traffic Modeling Module (300) Artificial Neural Network 2 (401) is being created. As a result of Artificial Neural Network 2 (401) calculations POP Assignment Lists 402 are determined for each P0P.

Claims (1)

REQUESTS . The invention is a performance enhancing System in Content Delivery Networks (CDN) and its feature is; Data Collection Module (1001), which collects data from CDN's PoP, Data Pre-Processing Module (200), which is used to make data suitable for use, Traffic Modeling Module (300) and Traffic Modeling Module (300), which creates traffic modeling by performing self-simular analysis on CDNs. It contains the P0P Assignment List Module (400) that generates the results and other modules. It contains the POP Raw Data (102) containing the POP Raw Data (102), the DNS Raw Data (103) containing the current DNS Data from the PoPs, and the POP Measurements (104) where the CPU, RAM, Disk Input / Output data from the P0Ps are collected. - It is a Processing Module (200), its feature is; Bandwidth Data (201), POP Raw data (102], which is input to the Traffic Modeling Module (300) by processing Bandwidth Raw Data (101] It contains DNS Data (203) that is input to the PoP Assignment List Module (400) by processing DNS Raw data (103) to the additional POP Assignment List Module (400). 4. The Traffic Modeling Module (300) mentioned in Claim 1 and its feature is; Bandwidth Data (301) dividing Bandwidth Data (201) from Data Preprocessing Module (200) into appropriate time intervals, Self-Similarity Analysis (302) performing statistical analysis of Bandwidth Data divided into time segments (301) with Self-similarity method (302) . Training Data (303) separated from Existing Bandwidth (301) Data, Test Data (304) separated from Existing Bandwidth (301) data, Training Data (303) that makes traffic modeling with Seasonal Auto-Regressive Integrated Moving Average (SARIMA) method. Model Parameters (307), which determines the model parameters that give the least error of Artificial Neural Network 1 (306), SARIMA (305) and Artificial Neural Network 1 (306) using the made models (307) ), using Model Parameters (307) and Error Analysis with Bandwidth Test Data (304), determine whether the best method in traffic modeling will be SARIMA (305) or Artificial Neural Network-1”. By calculating the best traffic modeling results, POP Assignment List Module' What (400) sends as input Result (309). 5. It is the POP Assignment List Module (400) mentioned in claim 1, and its feature is; Neural Network-2 (401), Artificial Neural Network - 2 As a result of (401) calculations, it includes PoP Assignment Lists (. 6. Invention, Content Delivery Networks (CDNH is a performance enhancing method, its feature is from Data Collection Module (100) to Data Pre-Processing Module, (201) Band Transmission of “Raw Data of Width” (101), transmission of POP Raw Data (101) from Data Acquisition Module (100) to Data Pre-Processing Module (201), Data Pre-Processing from Data Acquisition Module (100) Forwarding DNS Raw Data (101) to Module (201), transmitting POP Measurements (from Data Acquisition Module), transmitting ready-to-use Bandwidth Data in Data Preprocessing Module (200) to Traffic Modeling Module (201) ( 300), forwarding of POP Data ready-to-use in Data Preprocessing Module (200), ready-to-use DNS in Data Preprocessing Module (200). Transmission of the “stone” (, Bandwidth Data (301) divided into appropriate time intervals and transmitting for Self-Similarity Analysis (302), Storing a part of the Self-Similar Analysis (302) data as Training Data (303), Self-Similar Analysis (302) Storing a part of the analyzed (303) data as Test Data (302), modeling the Training Data ( and Artificial Neural Network l (306) by constructing it, Model Parameters of the SARIMA method (305) and Artificial Neural Network 1 (306) results) 307), Model Parameters (307) are determined, transmitting the traffic model to the Best Model for calculating margins of error, transmitting the Bandwidth Test Data (304) to the Best Model (308) for calculating margins of error, , so that the most valid Traffic Modeling Result (309) is forwarded to Neural Network 2 (401) in the P0P Assignment List Module (400), so that the results of Neural Network 2 (401) generate the POP assignment lists to send (402) to It contains m steps.