The Time-shift streaming

For alleviate the effect of the initial node-joining procedures and unfinished publishing and re-publishing procedures, we only examine the blocks generated between 30 to 90 minutes of each trial. Fig. 4-4 shows the cache results at each node;

node index below 50 are results from the live streaming nodes, and node index above 50 are results from the time-shift streaming nodes. Note that there are nodes suffering from DHT failures, which makes them unable to publish their file availability on the DHT. In the trial without the information of the number of currently active nodes, each node caches 67.86 files in average, with standard deviation of 56.39. and in the trial with the information of the number of currently active nodes, each node caches 61.87 files in average, with standard deviation of 35.54. We can also see from the figure that with the help of the information of currently active nodes of the system, the caching responsibility is more evenly distributed among nodes in the system.

Figure 4-4 Cache Results

Fig. 4-5 depicts the distribution of the number of replicas of each cached files among all live streaming nodes and time-shift streaming nodes. The method without the information of currently active nodes provides much more files that have ten replicas on the DHT, but in this method, more files are first cached and then deleted with the publish/republish mechanism. Note that files with more replicas would have more records on the DHT. On the other hand, although the method with the information of currently active nodes also provides ten replicas for most files on the DHT, but since there are more files that have less owner information found on the DHT comparing to the method without the information of the number of currently active nodes in the

system. The republishing processes may need a longer time to stabilize, in order to provide the information of cached file owner status on the DHT.

Figure 4-5 Distribution of Replica Count

Table 4-2 The sources of time-shift streaming blocks in the first trial

Node Index TS 01 TS 02 TS 03 TS 04 TS 05 TS 06 TS 07 TS 08 TS 09 TS 10 TS 11 TS 12 TS 13 TS 14 TS 15 TS 16 From other nodes 3657 3484 4122 1187 3441 5903 4035 5816 5539 5788 3652 4999 1729 2614 2995 2940

From Server 72 116 122 1413 100 189 96 276 310 307 107 64 6 38 89 489

Failed to Get 72 114 90 10 100 186 91 97 220 193 86 61 3 38 89 70

Emergency 0 2 26 1381 0 3 5 19 90 114 21 3 3 0 0 279

No Owner 0 0 6 22 0 0 0 160 0 0 0 0 0 0 0 140

Table 4-3 The sources of time-shift streaming blocks in the second trial time-shift streaming nodes in each trial, respectively. With the help of the server as an emergency handling node, both trials achieves 100% continuity index. In the first trial, the TS nodes had received a total of 65695 blocks, and 61901 (94.22%) of them were served by peer nodes. 3794 blocks were supported by the providing server, where 1946 of them were emergency handling, 1520 of them were unable to get from peers, and 328 of them have no file owner.Most of the requests to the providing server were by node TS 04, that suffered from a temporary DHT failure, and thus during that failure period, all blocks were supported by the server. In the second trial, the TS nodes had received a total of 59033 blocks, and 57343 (97.14%) of them were served by peer nodes, or. 1690 blocks were supported by the providing server, where 1044 of them were emergency handling, 522 of them were unable to get from peers, and 123 of them have no file owner. The reason of failing to get from peers was because just after the time-shift node acquires the owner list of its wanted file from the DHT, one of the owners in the list detects there is more than 10 replicas of that file in the system and deletes the file it cached, and thus the requests to the node that no longer has the file will all fail. Fig 4-8 depicts the distribution of each node’s contribution to the time-shift streaming nodes, node index below 50 are live streaming nodes, and node index above 50 are time-shift streaming nodes, and we can see the load is distributed through the nodes by the random

algorithm, and each node’s distribution is basically follows the number of its cached files, as shown in Fig. 4-4.

Figure 4-6 The Distribution of Nodes' Contribution to Time-shift Streaming Nodes

Chapter 5

Conclusion and Future Work

In this thesis, we had implemented a P2P live/time-shift streaming system and presented two distributed cache management strategies for time-shift video segments to cache a desired number of replicas based on the DHT knowledge and publishing/republishing mechanisms. We also studied the performance of the system on PlanetLab. Our experiment results show the feasibility of live/time-shift systems. In our experiments, the live streaming part achieved a startup delay of 16 seconds, an end-to-end delay of 120 seconds and a continuity index over 98%. Moreover, for the time-shift part, with the streaming server as an emergency handler, it achieved a continuity index of 100%, with over 94% of the streaming data were from P2P peers.

Our proposed caching strategies effectively distribute the load of storing the time-shift contents and provide ten replicas for most files.. The information of the number of currently active nodes in the system also helps in distributing the load for storing the time-shift contents more evenly among the nodes, as the standard deviation of the number of files cached on the nodes was reduced. .

However, in this implementation, publishing on the DHT has synchronization issues. Although random back-off publishing/republishing may alleviate this problem, collisions still occur, which lead to the difference between the owner lists on the DHT and the caching status in reality, and thus lowers the effectiveness of the cached files for time-shift viewers. More investigation is need on this DHT issue and larger experiments of the system would provide more insightful knowledge on P2P time-shift streaming services.

References

[1] F. Douglis and M.F. Kaashoek, “Scalable Internet Services,” IEEE Internet Computing, vol. 5, no. 4, 2001, pp. 36–37.

[2] Vakali, A.; Pallis, G., “Content delivery networks: status and trends” IEEE Internet Computing, vol. 7, no. 6, 2003, pp. 68-74.

[3] Xinyan Zhang, et al., “CoolStreaming/DONet: a data-driven overlay network for peer-to-peer live media streaming” INFOCOM 2005. 24th Annual Joint Conference of the IEEE Computer and Communications Societies. Proceedings IEEE, vol. 3, pp.

2102-2111. Mar. 2005

[4] Li Zhao, et al., “Gridmedia: A Practical Peer-to-Peer Based Live Video Streaming System” Multimedia Signal Processing, 2005 IEEE 7th Workshop on, pp. 1-4. Nov.

2005

[5] Bo Li, et al., “Inside the New Coolstreaming: Principles, Measurements and Performance Implications” INFOCOM 2008. The 27th Conference on Computer Communications. IEEE, pp. 1031-1039, Apr. 2008

[6] V. N. Padmanabhan, et al., “Distributing streaming media content using cooperative networking,” in Proc. 12th international workshop on Network and operating systems support for digital audio and video, pp. 177-186. Apr. 2002.

[7] M. Castro, et al., “Splitstream: High-bandwidth content distribution in a cooperative environment,” in Proc. nineteenth ACM symposium on Operating systems principles, pp. 292-303. Oct. 2003.

[8] Venkataraman, V. ; Yoshida, K. ; Francis, P., “Chunkyspread: Heterogeneous Unstructured Tree-Based Peer-to-Peer Multicast” Network Protocols, 2006. ICNP '06.

Proceedings of the 2006 14th IEEE International Conference on, pp 2-11. Nov. 2006 [9] Yang Guo, et al., “P2Cast: Peer-to-peer Patching Scheme for VoD Service”

Multimedia Tools and Applications, vol. 33, pp. 109-129, 2007

[10] Do, T.T. ; Hua, K.A. ; Tantaoui, M.A., “P2VoD: providing fault tolerant video-on-demand streaming in peer-to-peer environment” Communications, 2004 IEEE International Conference on, vol. 3, pp. 1467-1472, Jun. 2004

[11] Yi Cui ; Baochun Li ; Nahrstedt, K., “oStream: asynchronous streaming multicast

in application-layer overlay networks” Selected Areas in Communications, IEEE Journal on,

vol. 6, no. 1, Jan. 2004

[12] Dana, C. et al., “BASS: BitTorrent Assisted Streaming System for Video-on-Demand” Multimedia Signal Processing, 2005 IEEE 7th Workshop on, pp.

1-4. Nov.2005

[13] Yang Guo et al., “PONDER: Performance Aware P2P Video-on-Demand Service”

Global Telecommunications Conference, 2007. GLOBECOM '07. IEEE, pp. 225-230, Nov. 2007

[14] Deshpande, S. ; Noh, J.,“P2TSS: Time-shifted and live streaming of video in peer-to-peer systems” Multimedia and Expo, 2008 IEEE International Conference on, pp.649-652. Jun. 2008

[15] Hecht, F.V. et al. “LiveShift: Peer-to-Peer Live Streaming with Distributed Time-Shifting” Peer-to-Peer Computing , 2008. P2P '08. Eighth International Conference on, pp. 187-188, Sept. 2008

[16] Gallo, D. et al. “A Multimedia Delivery Architecture for IPTV with P2P-Based Time-Shift Support” Consumer Communications and Networking Conference, 2009.

CCNC 2009. 6th IEEE, pp. 1-2. Jan. 2009

[17] P. Maymounkov and D. Mazi`eres, “Kademlia: A peerto- peer information system based on the XOR metric.” Electronic Proceedings for the 1st International Workshop on Peer-to-Peer Systems, Mar. 2002

[18] Plan-x, http://www.thomas.ambus.dk/plan-x/routing/

[19] VideoLAN – VLC Media Player, http://www.videolan.org/vlc/

[19] PlanetLab, http://www.planetlab.org