Network Management and Analysis Lecture 3 PDF

Menoufia University Faculty of Computers and Information Department of Information Technology Network Management and Analysis Network Management and Analysis course. This will be an overview of the different topics that we will cover in this course. by Dr /Amina Mohamed Agenda Lec...

Menoufia University Faculty of Computers and Information Department of Information Technology Network Management and Analysis Network Management and Analysis course. This will be an overview of the different topics that we will cover in this course. by Dr /Amina Mohamed Agenda Lecture 3  Multimedia Network Application: Types Multimedia network application types: 1. Streaming stored audio/video 2. Conversational voice/video-over-IP 3. Streaming live audio/video Streaming Stored Video: Content Distribution Networks Streaming distribution: YouTube & Netflix, distribute hundreds of millions video every day. Single streaming data center approach. High possibility of low end-to-end throughput, why? Cost due to sending the same popular videos over and over again. Single point of failure. Content Distribution Networks (CDNs) approach. A CDN manages servers in multiple geographically distributed locations Direct each user request to the best CDN location. 4 Streaming Stored Video: Content Distribution Networks Content Distribution Networks (CDN): Single datacenter approach. 5 Streaming Stored Video: Content Distribution Networks Content Distribution Networks (CDN) 6 Streaming Stored Video: Content Distribution Networks Content Distribution Networks (CDN) 7 Streaming Stored Video: Content Distribution Networks Examples: Private CDN: Owned by content provider Google CDNs: Private CDN, used by YouTube Third party CDN: Owned by other companies. Akamai’s CDN: Third-party CDN, used by Netflix. 8 Streaming Stored Video: GOOGLE’S NETWORK INFRASTRUCTURE Case study: Google CDN Eight “mega data centers” 6 in the United States and 2 in Europe Each data center have 100,000 servers Store dynamic content About 30 “bring-home” clusters Location distributed around the world near tier-1 ISP 100–500 servers serving static content, including YouTube videos Many hundreds of “enter-deep” clusters located within an access ISP Tens of servers 9 Streaming Stored Video: GOOGLE’S NETWORK INFRASTRUCTURE Case study: Google CDN Data: Datacenter components Dr. Tamer Fathy 10 Streaming Stored Video: GOOGLE’S NETWORK INFRASTRUCTURE Case study: Facebook CDN data center view 11 Streaming Stored Video: GOOGLE’S NETWORK INFRASTRUCTURE Case study: Google CDN data center 12 Streaming Stored Video: Cluster Selection Strategies Cluster selection strategy: It is a mechanism for dynamically directing clients to a server cluster or a data center within the CDN. Geographically closest strategy: Using commercial geo- location databases, CDN data center is chosen based on distance to client used local DNS server. Challenge: User may use local DNS server far from this location. 13 Streaming Stored Video: DNS service Top level domain DNS server 14  Streaming Stored Video: Redirection to CDN network Click on URL NetCine http://video.netcinema.com/6Y7B23V. ma At website network a1105.k ingcdn. com Return IP address of a1105.ki content ngcdn.co server m King CDN network 15  Streaming Stored Video: Content provider case studies YouTube the world’s largest video-sharing site Owned by Google and uses Google CDN. CDN selection strategy: Geographical-closest based on local DNS. Netflix 30% of the downstream U.S. Internet traffic in 2011. Use third party infrastructure. Uses DASH in streaming. CDN selection strategy: Ranking in manifest file. 16  Streaming Stored Video: Content provider case studies Content ingestion Content processing Uploading versions to the CDNs Third party CDN: Akamai, Limelight, Level-3 17 Agenda Voice-over-IP Limitations of the Best-Effort IP Service Removing Jitter at the Receiver for Audio Recovering from Packet Loss  Conversational Streaming audio/video Example of multimedia network applications: Skype 19  Voice-over-IP: Limitations of the Best-Effort IP Service Internet Telephony & Voice-over-IP (VoIP): Real-time conversational voice over the Internet. Limitations of the Best-Effort IP Service: Delivering packets are not guaranteed with bounded delay or loss percentage. Challenge: VoIP is sensitive to packet loss, delay, and jitter. 20  Voice-over-IP: Limitations of the Best-Effort IP Service Packet loss: UDP: Delivery is not guaranteed due to network congestion. TCP: Reliable but at packet loss: Retransmission is a source of end-to-end delay variation. Congestion control may reduce the sending rate. 21  Voice-over-IP: Limitations of the Best-Effort IP Service Packet loss: 1-20 percent of packet loss can be tolerated depending on how voice is encoded and transmitted. Skype uses UDP unless a user is behind a NAT or firewall that blocks UDP segments (in which case TCP is used). 22  Voice-over-IP: Limitations of the Best-Effort IP Service Example: Audio playback rate: 8000 byte/sec. Periodical sending period: 20 ms Using UDP, Each audio chunk size: 0.02 sec x 8000 byte/sec = 160 byte. Challenges: Each chunk will not have the same end to end delay. 23  Voice-over-IP: Limitations of the Best-Effort IP Service End-to-End Delay: Delay sources: Routers: Queuing, processing, and transmission. Links: Propagation delays. End-system processing delays. Delay period and effect < 150ms: Not perceived. 150ms – 400ms: acceptable but not ideal. > 400ms: no acceptable and delayed packets are dropped. 24  Voice-over-IP: Limitations of the Best-Effort IP Service Packet Jitter: It is the variation of end to end delay with time. End to Sender Receiver end Jitter delay 0 ms 80 ms 80 ms packets 1 1 20 ms 130 ms 110 ms 30 ms 2 2 40 ms 160 ms 120 ms 10 ms 3 3 60 ms 170 ms 110 ms -10 ms 3 3 Ideal end to end delay 80 ms 25  Voice-over-IP: Removing Jitter at the Receiver for Audio Jitter removal: Sequence numbers and timestamps: Time at which the chunk was generated at sender. playout delay at receiver: Using buffer as shown earlier. 26  Voice-over-IP: Recovering from Packet Loss A packet is lost either if it never arrives at the receiver or if it arrives after its scheduled playout time. Retransmitting a packet that has missed its playout deadline serves absolutely no purpose. loss recovery schemes: Forward error correction (FEC). 27  Voice-over-IP: Recovering from Packet Loss 1. Forward Error Correction (FEC): Add redundant information to reconstruct approximations or exact versions of lost packets. Mechanism 1: Generating XOR 3achunks redundant encoded to generate the 4thchunk chunk after every n chunks using exclusive OR operation. Recover 8 7 only 6 for 5 one loss 4 inside 3 2 the 1group. Chunks sent from source Drawbacks: transmission rate: increased by 1/n. Playout delay: Must wait to receive the entire group. 28  Voice-over-IP: Recovering from Packet Loss Mechanism 2: The second FEC mechanism is to send a lower resolution audio stream as the redundant information.. 29  Voice-over-IP: Recovering from Packet Loss Mechanism 2 (continue …) Recovery: Only one loss in situations of successive losses. Improve overall audio quality. We may attach more than one low quality chunks. Example of audio bit rates: Main quality PCM 64 Kbps, Low quality GSM 13 Kbps. Drawbacks: 30  Voice-over-IP: Recovering from Packet Loss 2. Interleaving: The sender resequences units of audio data before transmission. At sender 31  Voice-over-IP: Recovering from Packet Loss 2. Interleaving (continue …) At reciever It does not increase the needed bandwidth. Increases latency 32