[5] [6] 23 YouTube 31% [5]. HTTP 12% YouTube 1 CoreLab [2], PlanetLab [3] [4] (In-Network ) CDN(Content Delivery Network) CDN CDN CDN CDN CDN CDN 1 IS

Size: px

Start display at page:

Download "[5] [6] 23 YouTube 31% [5]. HTTP 12% YouTube 1 CoreLab [2], PlanetLab [3] [4] (In-Network ) CDN(Content Delivery Network) CDN CDN CDN CDN CDN CDN 1 IS"

ひさともうみのなか
7 years ago
Views:

1 THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS TECHNICAL REPORT OF IEICE. In-Network ando@nakao-lab.org, nakao@iii.u-tokyo.ac.jp Youtube In-Network YouTube 3 In-Network () (1) (2) (3) In-Network In-Network Abstract A Study of In-Network Cache for Redundancy Elimination at Network Edges. Shogo ANDO and Akihiro NAKAO The University of Tokyo, Hongo, Bunkyo-ku, Tokyo, Japan ando@nakao-lab.org, nakao@iii.u-tokyo.ac.jp Network traffic conveying large content such as YouTube video steadily increasing in recent years now represents a very high proportion of the entire traffic flowing across the Internet. In addition, it is observed that such traffic transferring video streams includes high redundancy, since many users often access the same popular content at the same time. In this study, in order to eliminate redundancy near the edge of the network, using the traffic traces captured at a campus network gateway, we analyze YouTube access logs and replay the traffic under the following three scenarios of In-Network caching (enabling packet cache on routers and switches), (1) packet cache enabled at a campus gateway, (2) packet cache deployed at the very edge of the campus network, and (3) packet cache embedded in every router within the campus network. We evaluate how we can eliminate redundancy from the traffic flowing into the edge network through these three types of In-Network caching and show that we can reduce duplicate not only from in-coming traffic but also from the traffic within the campus network. Key words Redundancy Elimination, Network Cache, Traffic Engineering, Content Oriented Network 1. YouTube P2P 36-41% [1] 1

Shogo ANDO and Akihiro NAKAO The University of Tokyo, 7 3 1 Hongo, Bunkyo-ku, Tokyo, 113 0033 Japan E-mail: ando@nakao-lab.org, nakao@iii.u-tokyo.ac.

2 [5] [6] 23 YouTube 31% [5]. HTTP 12% YouTube 1 CoreLab [2], PlanetLab [3] [4] (In-Network ) CDN(Content Delivery Network) CDN CDN CDN CDN CDN CDN 1 ISP In-Network In-Network YouTube In-Network 6 ( 1) ( 2) ( 3) ( ) 2 (a) In-Network ( 1) ( 3) 33% In-Network ( 1) 0% ( 2) 18% ( 3) 21% (b) In-Network ( 1) YouTube 291.5GB ( 2) (291.5GB / ) 148.5MB ( 3) (291.5GB / ) 127.1MB ( 1) ( 2)

CDN CDN CDN CDN CDN 1 ISP In-Network In-Network 1 1 23 YouTube In-Network 6 ( 1) ( 2) ( 3) ( ) 2

3 2. 1 YouTube [5] GB (All Traffic) (Unique Traffic) 291GB (Redundant Traffic) 174GB YouTube 31% 3 50% 30 80% % 10 40% Cumulative Traffic (Gbyte) CDF All Traffic Unique Traffic Redundant Traffic Duplication Degree (times) 0.1 Time that elapese before next same video access Elapsed Time (minutes) CCDF The Number of Times Same Video Views 4 YouTube Yucheng CONIC [8] HTTP 25-50% Ashok SmartRE [9] ISP IP Jae Net-Serv [7] Active Netwroking MicroCDN JAVA Active Netwroking CDN MicroCDN In-Network IP CDN P2P YouTube 3 3. Zink YouTube [6] Proxy P2P In-Network 5 1 3

1 Time that elapese before next same video access 0 0 200 400 600 800 1000 1200 1400 Elapsed Time (minutes) CCDF 0.1 0.01 0.

4 5 In-Network 6-(1) 6-(2) In-Network 6-(3) In-Network ( 1) ( 2) ( 3) YouTube In-Network In-Network In-Network 1 In-Network In-Network In-Network In-Network ( 1) In-Network ( 1) In-Network ( 2) In-Network ( 3) In-Network YouTube YouTube ID HTTP In-Network ( 1) ( 2) 4

In-Network In-Network ( 1) In-Network ( 1) In-Network ( 2) In-Network

5 ( 3) 3 In-Network 3 LRU (Least Recently Used) (a)in-network 3 (b) [5] YouTube :40:10 JST :45:18 JST HTTP HTTP 4,452GB YouTube 503GB YouTube 503GB Range [10] 465GB 1, ,118 7 x y CCDF 90% 512MB 50% 44MB CCDF hour traffic arriving at each IP hour traffic arriving at each IP YouTube 1 NAT 1 IP 1 IP 1 YouTube 2 1 Web Web 1 CDN YouTube 1 CDN HTTP id itag ,963 traceroute traceroute Class B 3 prefix IP 256 1, Class B IP GB 0.125GB, 0.25GB GB (a) 5. 1 (b) x y 5

50% 44MB CCDF 10 0 10-1 10-2 10-3 10-4 23 hour traffic arriving at each IP 10-4 10-3 10-2 10-1 10 0 10 1 10 2 10 3 10 4 10 5 23 hour traffic arriving at each IP 7

6 ( 2) 18.1% ( 3) 21.9% ( 3) In-Network ( 3) ( 1) 33.0% 2 In-Network ( 3) ( 2) 20.7% In-Network ( 3) ( 1) ( 2) 1 In-Network 1: 33.0% 0% 2: 20.7% 18.1% 3: 33.0% 21.9% 10 x y In-Network ( 1) 2 [5] 31.0% In-Network ( 1) GB ( 2) MB ( 3) MB 256 6

6 10 9 1 In-Network ( 3) ( 1) ( 2) 1 In-Network 1: 33.0% 0% 2: 20.7% 18.

7 2 ( 1) ( 2) ( 3) 2 In-Network 1: 33.0% 0% 2: 12.1% 12.1% 3: 11.7% 11.6% ( 2) ( 3) 5. 1 ( 3) ( 2) 2 1 ( 3) MB ( 2) MB 2 ( 3) ( 2) 465GB 1GB 9 10 ( 3) ( 2) ( 3) ( 2) ( 3) ( 2) 5. 3 In-Network ( 1 2 3) ( 2 3) 7 3 (1) (2) ( 1) ( 2) 2 ( 3) YouTube 291.5GB 291.5G 145.8MB MB In- Network 2 In-Network In-Network In-Network 7

5 MB 2 ( 3) ( 2) 465GB 1GB 9 10 ( 3) ( 2) ( 3) ( 2) ( 3) ( 2) 5. 3 In-Network 5. 1 5.

8 3 () In-Network 33.0% 12.1% 14.9% 14.7% In-Network In-Network 2 In-Network ( 1) ( 2) ( 3) In-Network ( 1) ( 3) In-Network ( 3) In-Network ( 3) ( 1) ( 2) ( 1) ( 2) In-Network ( 1) ( 2) 2 ( 3) ( 1) ( 2) ( 3) ( 3) 6. 2 In-Network In-Network YouTube In-Network Click [11] [1] ns341/ns525/ns537/ns705/cisco VNI Usage WP.pdf. [2] Akihiro Nakao, Ryota Ozaki, Yuji Nishida, CoreLab: An Emerging Network Testbed Employing Hosted Virtual Machine Monitor, ACM ROADS [3] B. Chun, D. Culler, T. Roscoe, A. Bavier, L. Peterson, M. Wawrzoniak, M. Bowman. PlanetLab: an overlay testbed for broad-coverage services. SIGCOMM Comput. Commun. Rev., 33(3):3-12, [4]., Vol.94 No.5 pp , [5],,.,, Vol. 110, No. 448, NS , pp , [6] M. Zink, K. Suh, Y. Gu, and J. Kurose. Characteristics of youtube network traffic at a campus network - measurements, models, and implications. Comput. Netw., 53(4), [7] Jae Woo Lee, Roberto Francescangeli, Jan Janak, Suman Srinivasan, SalmanAbdul Baset, Henning Schulzrinne, Zoran Despotovic, Wolfgang Kellerer, Net-Serv: Active Networking 2.0, IEEE ICC 2011, Kyoto, Japan, June [8] Yucheng Zhu, Maoke Chen, Akihiro Nakao, CONIC: Content-Oriented Network with Indexed Caching, Global Internet, INFOCOM IEEE Conference on Computer Communications Workshops, [9] Ashok Anand, Vyas Sekar, and Aditya Akella. SmartRE: an architecture for co-ordinated network-wide redundancy elimination. SIGCOMM Comput. Commun.Rev., 39(4):87-98, [10] R. Fielding, J. Gettys, J. Mogu, H. Frystyk, L. Masinter, P. Leach, T. Berners-Lee Hypertext Transfer Protocol HTTP/1.1. Network Working Group. RFC2616, June [11] Click Modular router. 8

2 In-Network In-Network YouTube In-Network Click [11] [1] http://www.cisco.com/en/us/solutions/collateral/ ns341/ns525/ns537/ns705/cisco VNI Usage WP.pdf.

opentag_nv.pptx

opentag_nv.pptx OpenTag, This work is supported by Ministry of Internal Affairs and Communications of the Japanese Government. Agenda 1. Introduction 2. Design 3. Implementation 4. Evaluation 5. Conclusion 1. Introduction