SLIDE 1
Popularity Prediction of Facebook Videos for Higher Quality Streaming
1
Videos are Central to Facebook
8 billion views per day
2
Popularity Prediction of Facebook Videos for Higher Quality - - PowerPoint PPT Presentation
Popularity Prediction of Facebook Videos for Higher Quality - - PowerPoint PPT Presentation
Popularity Prediction of Facebook Videos for Higher Quality Streaming Linpeng Tang Qi Huang , Amit Puntambekar Ymir Vigfusson , Wyatt Lloyd , Kai Li 1 Videos are Central to Facebook 8
SLIDE 2
SLIDE 3
Workflow of Videos on Facebook
3
Original
Streaming Video Engine
CDN
Encoded ABR Streaming Upload
Backend Storage
ABR streams the best quality version of the video that fits!
Intensive processing needed to create multiple video versions for ABR streaming
SLIDE 4
Better Video Streaming from More Processing
- Better compression at the same quality
- QuickFire: 20% size reduction using
20X computation
- More users can view the high quality versions
4
BaQdwith Video Quality Alice Bob
SLIDE 5
Better Video Streaming from More Processing
- Better compression at the same quality
- QuickFire: 20% size reduction using
20X computation
- More users can view the high quality versions
5
BaQdwith Video Quality Alice Bob
Better Compression
SLIDE 6
How to apply QuickFire for FB videos
- Infeasible to encode all videos with QuickFire
– Increase by 20X the already large processing fleet
- High skew in popularity
– Reap most benefit with modest processing?
6
SLIDE 7
Opportunity: High Skew in Popularity
- Access logs of 1 million videos randomly sampled by ID
- Watch time: total time users spent watching a video
7
100 101 102 103 104 105 106
Video rank
0.0 0.2 0.4 0.6 0.8 1.0
Cumulative watch time ratio
SLIDE 8
100 101 102 103 104 105 106
Video rank
0.0 0.2 0.4 0.6 0.8 1.0
Cumulative watch time ratio
Opportunity: High Skew in Popularity
- We can serve most watch time even with a small
fraction of videos encoded with QuickFire
- Can we predict these videos for more processing?
8
80%+ watch time
SLIDE 9
CHESS Video Prediction System
- Popularity prediction is important for higher
quality streaming
– Direct encoding on videos with the largest benefit
- Goal of CHESS video prediction system
– Identify videos with highest future watch time – Maximize watch-time ratio with budgeted processing
9
SLIDE 10
CHESS Video Prediction System
10
Streaming Video Engine
CDN
Backend Storage CHESS-VPS
Predicted Popular Videos Social signals
Facebook Graph Serving System
Access logs
SLIDE 11
CHESS Video Prediction System
11
Streaming Video Engine
CDN
Backend Storage
Predicted Popular Videos QuickFire Encoded
CHESS-VPS
Original Social signals
Facebook Graph Serving System
Access logs
SLIDE 12
Social signals
Facebook Graph Serving System
Access logs
CHESS Video Prediction System
12
Streaming Video Engine
CDN
Backend Storage
Predicted Popular Videos QuickFire Encoded
CHESS-VPS
Original
Serving QuickFire-encoded versions!
SLIDE 13
Requirements of CHESS-VPS
- Handle working set of ~80 million videos
- Generate new predictions every few minutes
- Requires a new prediction algorithm: CHESS!
13
SLIDE 14
CHESS Key Insights
- Efficiently model influence of past accesses as
the basis for scalable prediction
- Combine multiple predictors to boost accuracy
14
SLIDE 15
Efficiently model past access influence
- Self exciting process
– A past access makes future accesses more probable, i.e. provides some influence on future popularity
15
1 2 3 4 5
7ime
1 2 3 4 5 6 7 8
InIluence
1 2 3 4 5
7ime
1 2 3 4 5 6 7 8
InIluence
Influence of past accesses
SLIDE 16
1 2 3 4 5
7ime
1 2 3 4 5 6 7 8
InIluence
Efficiently model past access influence
- Self exciting process
– A past access makes future accesses more probable, i.e. provides some influence on future popularity – Prediction: sum up total future influence of all past accesses
16
Total future influence now
SLIDE 17
Efficiently model past access influence
- Influence modeled with kernel function
- Power-law kernel used by prior works
– Provides high accuracy – Scan all past accesses, O(N) time/space not scalable
17
1 2 3 4 5
TLPe
0.0 0.5 1.0 1.5 2.0
InIluence
3ower Law
y = (x + β)−α
SLIDE 18
Efficiently model past access influence
- Influence modeled with kernel function
- Power-law kernel used by prior works
- Key insight: use exponential kernel for scalability
18
1 2 3 4 5
TLPe
0.0 0.2 0.4 0.6 0.8 1.0
InIluence
3ower Law ExponentLal
y = exp(−x/w)
y = (x + β)−α
SLIDE 19
Efficiently model past access influence
- Self exciting process with the exponential kernel
19
˜ F (t) = x w + exp ✓−(t − u) w ◆ ˜ F (u)
Current Access Watch-time + Previous Prediction Exponential Decay x
SLIDE 20
Efficiently model past access influence
- Single exponential kernel is less accurate than
power-law kernel
– 10% lower watch time ratio
- O(1) space/time to maintain
20
Single exponential kernel is less accurate yet scalable
SLIDE 21
Combining Efficient Features in a Model
21
- Key insight: maintain multiple exponential kernels
- O(1) space/time
Exp Exponential ke kernels ls Mod Modeled by
Combining multiple exponential kernels is as accurate as a power-law kernel
Time WaWch Time
Ac Actual acces access pa pattern
SLIDE 22
Combining Efficient Features in a Model
22
Social signals further boosts accuracy
Future Popularity
Neural Network
Raw features
Multiple Kernels Directly-used Features likes comments shares
- wner likes
video age Past access watch-time
SLIDE 23
CHESS Video Prediction System
23
Aggr Aggregated top videos Aggr Worker1 Worker2 Worker3 Worker4 Prediction workers Shard1 Shard2 Shard3 Shard4 Access logs Streaming Model Model NN Models Client Client
SLIDE 24
Evaluation
- What is the accuracy of CHESS?
- How do our design decisions on CHESS affect its
accuracy and resource consumption?
- What is CHESS’s impact on video processing and
watch time ratio of QuickFire?
24
SLIDE 25
Evaluation
- What is the accuracy of CHESS?
- How do our design decisions on CHESS affect its
accuracy and resource consumption?
- What is CHESS’s impact on video processing and
watch time ratio of QuickFire?
25
SLIDE 26
Metrics
- Watch time ratio
– Ratio of watch time from better encoded videos – Directly proportional to benefits of better encoding
- Processing time
26
SLIDE 27
Metrics
- Watch time ratio
– Ratio of watch time from better encoded videos – Directly proportional to benefits of better encoding
- Processing time (infeasible to encode all videos)
– Video length processing time – Video length ratio ≈ computation overhead
27
∝
SLIDE 28
CHESS is Accurate
28
- Vary video length ratio (proxy for processing overhead)
- Observe watch time ratio of better encoded videos
10-6 10-5 10-4 10-3 10-2 10-1 100
Video lengWh raWio
0.0 0.2 0.4 0.6 0.8 1.0
WaWch Wime raWio
SLIDE 29
CHESS is Accurate
29
10-6 10-5 10-4 10-3 10-2 10-1 100
Video lengWh raWio
0.0 0.2 0.4 0.6 0.8 1.0
WaWch Wime raWio
IniWial(1d) (CAC0'10)
- Initial(1d): initial watch time up to 1 day after upload
SLIDE 30
CHESS is Accurate
30
10-6 10-5 10-4 10-3 10-2 10-1 100
Video lengWK rDWio
0.0 0.2 0.4 0.6 0.8 1.0
WDWcK Wime rDWio
6(I60IC (.'''15) IniWiDl(1d) (CAC0'10)
- Initial(1d): initial watch time up to 1 day after upload
- SESIMIC: handcrafted power-law kernel
SLIDE 31
10-6 10-5 10-4 10-3 10-2 10-1 100
VidHo lHngWK rDWio
0.0 0.2 0.4 0.6 0.8 1.0
WDWcK WimH rDWio
CH(66 6(I60IC (.'''15) IniWiDl(1d) (CAC0'10)
CHESS is Accurate
31
- Initial(1d): initial watch time up to 1 day after upload
- SESIMIC: handcrafted power-law kernel
CHESS provides higher accuracy than even the non-scalable state of the art
SLIDE 32
10 20 30 40 50 60
3URcHssing C38 (%)
0.3 0.4 0.5 0.6 0.7 0.8 0.9
WaWch 7iPH 5aWiR
CH(66 2wnHU OikHs
CHESS Reduces Encoding Processing
- Predict on whole Facebook video workload in real-time
- Sample 0.5% videos for actual encoding
32
CHESS reduces CPU by 3x (54% to 17%) for 80% watch time ratio
SLIDE 33
Related Work
33
Popularity Prediction Video QoE Optimization Caching
Hawkes'71, Crane'08, Szabo'10, Cheng'14, SEISMIC'15 Liu'12, Aaron'15, Huang'15, Jiang'16, QuickFire'16 LFU‘93, LRU’94, SLRU‘94, GDS’97, GDSF‘98, MQ’01 CHESS is scalable and accurate Optimize encoding with access feedback Identify hot items to improve efficiency
SLIDE 34
Conclusion
- Popularity prediction can direct encoding for higher
quality streaming
- CHESS: first scalable and accurate popularity predictor
– Model influence of past accesses with O(1) time/space – Combine multiple kernels & social signals to boost accuracy
- Evaluation on Facebook video workload
– More accurate than non-scalable state of the art method – Serve 80% user watch time with 3x reduction in processing
34