Proxy Caching Mechanism for Multimedia Playback Streams in

Proxy Caching Mechanism for Multimedia Playback Streams in the Internet R. Rejaie, M. Handley, H. Yu, D. Estrin USC/ISI http: //netweb. usc. edu/reza/ WCW’ 99 April 1, 1999 USC INFORMATION SCIENCES INSTITUTE 1

Motivation n Rapid growth in deployment of realtime streams(audio/video) over the Internet n Goals Š Š Maximize the quality of the delivered stream Minimize startup latency Low-latency VCR-functionality Minimize the load on the server & the network USC INFORMATION SCIENCES INSTITUTE 2

Outline n n An End-to-end Architecture Multimedia Proxy Caching Conclusion Future Directions USC INFORMATION SCIENCES INSTITUTE 3

Streaming Applications in Best-effort Networks (The Internet) n End-to-end congestion control is crucial for stability, fairness & high utilization Š n Results in variable transmission rate Streaming applications require constant average consumption rate Ô Streaming USC INFORMATION SCIENCES INSTITUTE applications should be quality adaptive 4

Quality Adaptation(QA) Buffering only absorb shortterm variations BW(t) n Long-lived session could result in buffer overflow or underflow n QA is complementary for buffering Ô Adjust the quality(rate) with long-term variations n Ô Layered USC INFORMATION SCIENCES INSTITUTE Time framework 5

The End-to-end Architecture Error Control Client Quality Cong. Adaptation Control Buffer Manager Archive Transmission Buffer Internet Server Acker Playback Buffer Manager Decoder Adaptation Buffer Data path Control path USC INFORMATION SCIENCES INSTITUTE 6

Limitation n Delivered quality is limited to the average bandwidth between the server and client Client ISP Internet Solutions: Š Š Mirror servers Proxy caching Server Quality(layer) n L 4 L 3 L 2 L 1 L 0 Time USC INFORMATION SCIENCES INSTITUTE 7

Multimedia Proxy Caching n Assumptions Š Client Proxy can perform: Client Proxy – End-to-end congestion ctrl – Quality Adaptation Internet n Goals of proxy caching Š Š Š Improve delivered quality Low-latency VCR-functions Natural benefits of caching USC INFORMATION SCIENCES INSTITUTE Server 8

Challenge n Cached streams have variable quality Played back stream Ô Layered organization provides opportunity for adjusting quality Quality (no. active layers) Stored stream L 4 L 3 L 2 L 1 L 0 Time USC INFORMATION SCIENCES INSTITUTE 9

Issues n Delivery procedure Š Š n Relaying on a cache miss Pre-fetching on a cache hit Replacement algorithm Š Š Determining popularity Replacement pattern USC INFORMATION SCIENCES INSTITUTE 10

Cache Miss Scenario n n Stream is located at the original server Playback from the server through the proxy Proxy relays and caches the stream No benefit in a miss scenario USC INFORMATION SCIENCES INSTITUTE Client Proxy Internet Server 11

Cache Hit Scenario n Playback from the proxy cache Š Š n Lower latency May have better quality! Available bandwidth allows: Š Š Lower quality playback Higher quality playback USC INFORMATION SCIENCES INSTITUTE Client Proxy Internet Server 12

Lower quality playback n n Missing pieces of the active layers are pre-fetched ondemand Required pieces are identified by QA Results in smoothing Quality (no. active layers) n Pre-fetched data Played back stream Stored stream L 4 L 3 L 2 L 1 L 0 Time USC INFORMATION SCIENCES INSTITUTE 13

Higher quality playback n Pre-fetch higher layers on-demand Pre-fetched data Played back Stream n n n Pre-fetched data is always cached Must pre-fetch a missing piece before its playback time Tradeoff USC INFORMATION SCIENCES INSTITUTE Quality (no. active layers) Stored stream L 4 L 3 L 2 L 1 L 0 Time 14

Replacement Algorithm n Goal: converge the cache state to optimal Š Variation in quality inversely depends on Client Proxy Average quality of a cached stream depends on – popularity – average bandwidth between proxy and recent interested clients Š Client Internet Server – popularity USC INFORMATION SCIENCES INSTITUTE 15

Popularity n n Number of hits during an interval User’s level of interest (including VCR-functions) whit = Playback. Time(sec)/Stream. Length(sec) n Potential value of a layer for quality adaptation Š n Calculate whit on a per-layer basis Layered encoding guarantees monotonically decrease in popularity of layers USC INFORMATION SCIENCES INSTITUTE 16

Replacement Pattern n n Cached segment Fine-grain Coarse-grain n Multi-valued replacement decision for multimedia object Coarse-grain flushing Š on a per-layer basis Quality(Layer) Fine-grain flushing Š on a per-segment basis Time USC INFORMATION SCIENCES INSTITUTE 17

Conclusion n End-to-end architecture for delivery of qualityadaptive multimedia streams Š n Congestion control & Quality adaptation Proxy caching mechanism for multimedia streams Š Š Pre-fetching Replacement algorithm Ô State of the cache converges to the optimal USC INFORMATION SCIENCES INSTITUTE 18

Future Directions n Extensive simulation(using VINT/ns) Š n n e. g. access pattern, the bandwidth distribution Exploring other replacement patterns Chunk-based popularity function USC INFORMATION SCIENCES INSTITUTE 19

Alternative Replacement Algorithm Goal: to cache popular portion of each stream n Keep track of per-chunk popularity n Identify a victim chuck n Apply the same replacement pattern within the victim chunk n USC INFORMATION SCIENCES INSTITUTE 20