컴퓨터 네트워크 ch2(6)

Video streaming and content distribution networks (CDNs)

---

 

Internet Video

• video tarffic : major consumer of Internet bandwidth
  • Netflix, Youtube : 37%, 16% of downstream residential ISP traffic
  • ~ 1B YouTube users, ~ 75M Netflix users
• challenge
  • scalibility - single mega-video server won't work
  • heterogenity
    • different users have different capabilities (e.g., wired versus mobile; bandwidth rich vershs bandwidth poor)
• solution : distributed, application-level infrastructure

 

Video encoding

• video : sequence of imags displayed at constant rate (e.g., 24 or 30 frames/sec)
• digital image : array of pixels
• coding : use redundancy within and between images to decrease number of bits encode image
  • spatial (within image) - 인접한 pixel 간의 압축
  • temporal (from one image to next) - frame 사이의 pixel간에 압축 이용

 

Video Encoding

• CBR (constant bit rate) : video encoding rate fixed
• VBR (variable bit rate) : video encoding rate changes as amount of spatial, temporal coding changes (압축 더 많이 할 수 있을 때는 더 함)
• examples:
  • MPEG 1 (CD-ROM) 1.5Mbps
  • MPEG2 (DVD) 3-6Mbps
  • MPEG4 (often used in Internet, < 1Mbps)

 

Streaming multimedia

• Traditional streaming service
  • Based on RTSP (Real-Time Streaming Protocol) [RTP, RTSP]
  • Complex : burden to a server
• 요즘 - UTP (UDP 기반)
• Session management and flow control

 

• Streaming over HTTP - (GET으로 받음) TCP 기반

한꺼번에 왕창받으면 bufferr가 많이 필요하다.

그러나 사용자들이 도중에 많이 중단하는데 이는 bandwidth낭비가 된다.

 

Progressive Downloading : 일부 chunk만 요구. play하다가 또 chunk를 요구한다.

이는 일부 bandwidth만 낭비한다.

 

 

 

Streaming multimedia : DASH

• DASH : Dynamic, Adaptive Streaming over HTTP
• DASH로 heterognity 해결!
• server :
  • divides video file into multiple chunks
  • each chunk stored, encoded at different rates - 압축을 고화질/저화질... 다양한 옵션이 가능
  • manifest file : provide URLs for different chunks - 어떤 rate를 가진 url이 무엇인지 기록
• client:
  • first requests the manifest file - MPEG에서는 MPD(Media Presentation Description)
  • periodically measures server-to-client bandwidth
  • consulting the manifest, requests one chunk at a time
  • choose maximum coding rate sustainable given current bandwidth (현재 availble한 bandwidth에 따라 chunk 요청)
  • can choose different coding rates at different points in time (depending on available bandwidth at time) - 화질이 나빠질 수는 있으나 끊기지 않는다!

 

Streaming multimedia : DASH

"inteligence" at client : client determines

- 언제 chunk를 요구할 것인가 (buffer starvation, or overflow does not occur)

- 어떤 encoding rate를 요구할 것인가 (higher quality when more bandwidth available)

- 어디서 chunk를 요구할 것인가 (can request from URL server that is "close" to client or has high available bandwidth)

 

 

 

 

Content Distribution Networks(CDNs)

Challenge : 사용자가 너무 많다

 

Option 1 : single, large "mega-server"

• single point of failure (centralized 단점)
• point of network congestion
• long path to distnat clients - the average end-to-end BW 보장되어야 하는데 bottleneck 발생
• multiple copies of video sent over outgoing link - 인기많은  비디오라서 동일한 traffic이 반복하여 link를 통해 전달될 수 있다. 가는 길이 겹쳐서 congestion 발생
• this solution doesn't scale

 

Opetion2 : store/serve multiple copies of videos of multiple geographically distributed sites (CDN)

• Enter deep : CDN 서버를 사용자와 가능한 가까이 설치
  • used by Akamai, 1700 locations
• Bring home : access networ 근방 POP에 큰 용량의 cluster를 smaller number(10's)만큼 설치하겠다
  • used by Limelight

 

CDNs which content is retrieved

• CDN : stores copies of content at CDN nodes
  • 여러 서버에 분산시켜 두고, original server에 manifest file을 기록한다. 
  • e.g. Netflix stores copies of MadMen
• subscriber requests content from CDN
  • 사용자 근처에서 content를 고르거나
  • 만약 그 길이 traffic이 많으면 다른 것을 고른다

 

OTT challenges : coping with a congested Internet

• from which CDN node to retrieve content?
• viewer behavior in presence of congestion?
• what content to place in which CDN node?
• Over the top (OTT)
  • A term used to refer to content providers that distribute streaming media as a standalone product directly to viewers over the Internet, by passing telecommunications, multichannel televison, and broadcast television platforms that traditionally act as a controller or distributor of such content.

 

 

CDN Content Access : a closer look

Bob(client) requests video http://netcinema.com/6Y7B23V 

video stored in CDN at http://KingCDN.com/NetC6Y&B23V 

 

2번에서 netcinema.com/6Y7B23V에 대한 IP주소 요청

redirection하여 가까운 쪽 서버에서 받아가도록 한다.

 

 

 

CDN Content Access : Netflix

- No DNS redirection

- Push caching : CDN서버에 어떤 영화를 둘 것인가! 어떤 지역에서, 지역 사람들이 특정 영화를 많이 본다고 할 때 이를 예측하여 traffic이 적을 때, 미리 영화를 server에 넘겨둠

 

 

CDN Content Access : Youtube

• 구글이 인수
• Use its private CDN to distribute YouTube Video
• Use DNS redirection : redirection 시킬 때, node selection algorithm으로 RTT(거리 traffic) 측정함 (load balance)
• Use push caching ; 자기 가까운 쪽에 요청하면 가까운 쪽 서버에 요청. 그 서버에 없으면 original server에서 받음 (마치 web cache와 비슷)
• MPEG-DASH 알고리듬
• HTTP streaming with no adaptive streaming
• The user manually selects a video version
• Currently using MPEG-DASH
• Cluster selection strategy
• Selects a cluster which has the lowest RTT while considering the load balance
• Kankan : 처음엔 P2P이지만, P2P로만 받을 때 문제가 생기면 CDN을 취한다.
  • Originally based on P2P, migrated to a hybrid CDN-P2P