All Roads Lead to Network Coding a perspective
- Slides: 18
All Roads Lead to Network Coding ——a perspective on the application of Network Coding in information processing Hui Zhang EMC Research China November 2010 © Copyright 2010 EMC Corporation. All rights reserved. 1
Network Coding Roadmap “Network Coding” Cite Frequency a 1 a S b 2 b 3 a b a+b 4 a+b T 2 T 1 © Copyright 2010 EMC Corporation. All rights reserved. 2
EMC’s Investment in Information Procession and Management © Copyright 2010 EMC Corporation. All rights reserved. 3
Agenda Background Network Coding Application Cases – Information Transmission: Video Streaming – Information Transmission: Video Conference – Information Storage: Raid Conclusion © Copyright 2010 EMC Corporation. All rights reserved. 4
Network Coding Application Case ——Video Streaming Live P 2 P Streaming – Not enough throughput – Large playout delay – Packet losses Random Linear Network Coding – – Make full use of throughput resource Reduce playout delay Increase robustness Ease protocol design © Copyright 2010 EMC Corporation. All rights reserved. 5
Random Linear Network Coding in P 2 P Streaming Segmentation of Data Packets Sequence – Original sequence in live P 2 P streaming needs to be partitioned into segments for encoding and decoding on individual segment. Transmission Process – Take one segment as an example, each node pushes data packets to its neighbors based on received buffer status packets from them. – Encoded data packets can be decoded when this segment is full. Otherwise, they can not be decoded. © Copyright 2010 EMC Corporation. All rights reserved. 6
The Advantage of Network Coding Network coding lets users have more opportunities to get the data packets they need, because coded data packets in the same segment are considered “equally”. © Copyright 2010 EMC Corporation. All rights reserved. 7 7
Network Coding Based Live P 2 P Streaming System Compared with the live P 2 P streaming system without network coding (No. NC), network coding (NC) based live P 2 P streaming system adds two modules to the system diagram. © Copyright 2010 EMC Corporation. All rights reserved. 8 8
Experiment Results (Ns. Lab, Tsinghua) Under simple scheduling and transmission protocol Network Coding reduce redundancy ratio while maintain high quality. 0. 5 0. 4 NC 0. 3 No. NC 0. 2 0. 1 0 NC No. NC 2 1. 1 3 4 5 6 Scenario 7 8 1. 1 1 NC 0. 7 No. NC Quality 0. 9 0. 8 2 3 9 GREEDY 1 4 5 6 Scenario 7 8 1 2 3 4 5 6 Scenario 7 8 9 © Copyright 2010 EMC Corporation. All rights reserved. NC No. NC 1. 1 0. 9 0. 8 NC 0. 7 2 3 No. NC 4 5 6 Scenario 7 8 9 RANDOM 1 0. 5 RANDOM 9 RAREST_FIRST 0. 6 0. 5 0. 4 0. 35 0. 3 0. 25 0. 2 0. 15 0. 1 0. 05 0 1 1 1 Quality RAREST_FIRST Redundancy Ratio 0. 6 Redundancy Ratio GREEDY 0. 5 0. 4 0. 35 0. 3 0. 25 0. 2 0. 15 0. 1 0. 05 0 Quality 0. 7 0. 9 0. 8 NC 0. 7 No. NC 0. 6 0. 5 1 2 3 4 5 6 Scenario 7 8 9 9
Agenda Background Network Coding Application Cases – Information Transmission: Video Streaming – Information Transmission: Video Conference – Information Storage: Raid Conclusion © Copyright 2010 EMC Corporation. All rights reserved. 10
Network Coding in Video Streaming n Video Conference vs. Video Streaming q q n Goal q q n Critical delay and buffer time Small communication scale Reduce Delay Maintain Throughput Example [Philip Chou, 2004] © Copyright 2010 EMC Corporation. All rights reserved. 11
Experiment Results (Ns. Lab, Tsinghua) Average Marxism Local Delay © Copyright 2010 EMC Corporation. All rights reserved. Marxism Global Local Delay 12
Agenda Background Network Coding Application Cases – Information Transmission: Video Streaming – Information Transmission: Video Conference – Information Storage: Raid Conclusion © Copyright 2010 EMC Corporation. All rights reserved. 13
Raid Redundant Array of Independent (Inexpensive) Disks A set of disk arrays treated as one logical disk Data are distributed over the disks Redundant capacity is used for parity allowing for data repair Scale up ” 3. 5 ” 5. 25 Scale out ” 10 ” 14 ” 3. 5 © Copyright 2010 EMC Corporation. All rights reserved. 14
Level Raid 0 Raid 1 Raid 2 Raid 3 Raid 4 Raid 5 Raid 6 Description block-level striping mirroring dedicated Hamming dedicated parity (byte) dedicated parity(block) distributed parity (1) distributed parity (2) Mini Disk 2 2 3 3 4 Space Efficiency 1 1/n 1 - 1/n ⋅ log 2(n+1) 1 -1/n 1 -2/n Fault Tolerance None n-1 disk 1 disk 2 disk Read good good Write good OK OK © Copyright 2010 EMC Corporation. All rights reserved. 15
RAID Higher throughput, higher transaction rate and Fault tolerance – Dominates today’s large storage systems Trade off between reliability and availability – Reliability : Is anything broken? – Availability: Is the system still available to the user? (Read, Write and Consistency( RAID-0 – No Protection, Best Performance, Least Cost RAID-1 – Best Protection, Good Performance, Most Expensive RAID-5 – Good Protection, OK Performance, Least Expensive Fault Tolerant RAID-6 – Better Protection, Worse Performance, More Expensive © Copyright 2010 EMC Corporation. All rights reserved. 16
Conclusion Big success both in academic and industry Future Directions – Theory enrichment – Application area – Cooperation with other coding © Copyright 2010 EMC Corporation. All rights reserved. a 1 a S b 2 b 3 a b a+b 4 a+b T 2 T 1 17
© Copyright 2010 EMC Corporation. All rights reserved. 18
- Apa itu selective coding
- Open coding
- Modern roads vs roman roads venn diagram
- Not all roads lead to heaven
- Many roads lead to the same destination
- History of qualitative research
- Coding dna and non coding dna
- 5 lead ecg placement colors
- Frequency dependence of dielectric constant
- Draw three noncollinear points j k and l
- Stairs one point perspective
- Silo perspective vs business process perspective
- Network coding
- Network coding
- Weichert lead network
- Ancient china han
- Roadside friction
- Advertising signs on public roads
- Disadvantages of kinetic roads