CACHING TO SUPPORT MULTI-CLASS CONTENT AND NODE MOBILITY FOR NAMED DATA NETWORK

• Main Author: Vidya Yovita, Leanna
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/57235
• Institution: Institut Teknologi Bandung

In an internet network (IP network), users who need a content will send their requests to certain addresses on the network, therefore requests for the same content repeatedly will cause unnecessary burden on the network and cause communication delays. To overcome this problem, starting in 2009, Jacobson et al. proposes a new content-based networking paradigm. This paradigm, called the Named Data Network, causes responses to content requests by users not only to be served by certain servers, but also to the nearest device that has the requested content. To support the Named Data Network concept, the NDN router nodes are equipped with content stores to store data (cache). Almost all of existing research on caching focuses on a single technique of determining the location, selecting content to be cached or deleted, or collaboration between content stores, where the modeling uses one or several services regardless of class. Most methods only differentiate the treatment of packages based on the content popularity. Meanwhile, currently there are contents that has different characteristics, such as real time entertainment, web page, social networking, real time communication, and others. This different character must be considered in determining the storage rules in the content store to provide appropriate performance for each content class and improve the overall performance of the system. A solution is needed, a caching algorithm that can adjust the content storage strategy according to the character of the content class so that it can provide optimal performance for each content class. This dissertation research was carried out to propose a new caching algorithm that paying attention to the content popularity, and also notice the various classes of content so that this caching algorithm can provide optimal performance for the Named Data Network, for each content class and the system. The proposed caching algorithm considers three strategies simultaneously: where to place content, what content to be stored, its portion and the content replacement in the content store. This research was carried out in 2 major stages. In the first stage, a new caching algorithm is created to accommodates content popularity and different treatment vi of various content classes, including the placement of content in network nodes, the selection of content and content deletion from the cache. In the second stage, a caching algorithm was developed. It accommodates changes in consumer demand patterns dynamically and the movement of nodes on NDN. The mathematical analysis is used as the basis of a formula for determining the proportion of cache for each content class in the caching algorithm 'Cache based on Popularity and Class'. In this dissertation, CAPIC was developed through 2 algorithms, namely Static-CAPIC and Dynamic-CAPIC. The two algorithms proposed in this dissertation, namely Static-CAPIC and Dynamic-CAPIC have their respective advantages and both have succeeded in providing improved performance compared to the combination of the two existing caching algorithms, LCD+Sharing. Static-CAPIC provides advantages because of its simple mechanism. This algorithm provides better performance than the combination of caching algorithm LCD and sharing. However, Static-CAPIC cannot accommodate dynamic changes in consumer demand. The Static-CAPIC algorithm was then further developed into Dynamic-CAPIC to accommodate changing conditions of consumer demand. With the dynamic proportion calculation formula in Dynamic-CAPIC, each NDN router will be able to calculate the cache proportion for each service class in real-time. The two algorithms proposed in this dissertation, Static-CAPIC and DynamicCAPIC provide a higher cache hit ratio of the system than the LCD+Sharing. Static-CAPIC provides a simple cache technique and provides a higher cache hit ratio than the combination of the two common techniques, namely LCD+Sharing. Dynamic-CAPIC, as a developed algorithm of Static-CAPIC provides the flexibility to change the proportion of cache based on the frequency of consumer requests in real-time. This technique provides an appropriate cache hit ratio and path stretch for each content class character, where the largest cache hit ratio for the first class is 130.7% greater than Static-CAPIC and 92.4% higher compared to LCD+Sharing, followed by third class, where the cache hit ratio is 10.98% higher than LCD+Sharing and smaller than Static-CAPIC and then the second class, where the cache hit ratio is 4.21% higher than static-CAPIC, and 6.95% lower than LCD+ share. The smallest path stretch is in the third class, followed by the second class and the first class. This algorithm also outperforms Static-CAPIC, and LCD+Sharing schemes in the total network cache hit ratio parameter, where Dynamic-CAPIC provides a cache hit ratio of 6.29% higher than static-CAPIC and 15.15% higher than LCD+Sharing. In addition, Dynamic-CAPIC is proven to be able to accommodate node movements and also large differences in demand rates for each class. The advantages of Dynamic-CAPIC are also increasingly visible in network conditions with a large number of demand levels. At request level 6, Dynamic-CAPIC provides a cache hit ratio of 10.5% greater than LCD+sharing and 11.7% than Static-CAPIC.