Buffer sizing and Video QoE Measurements at Netflix Bruce Spang , - - PowerPoint PPT Presentation

▶
buffer sizing and video qoe measurements at netflix
SMART_READER_LITE
LIVE PREVIEW

Buffer sizing and Video QoE Measurements at Netflix Bruce Spang , - - PowerPoint PPT Presentation

Nov 14, 2023 776 likes •1.17k views

Buffer sizing and Video QoE Measurements at Netflix Bruce Spang , Brady Walsh, Te-Yuan Huang, Tom Rusnock, Joe Lawrence, Nick McKeown December 2, 2019 How big should a buffer be? Too big: packets wait for too long Too small: too many packets

slide-1
SLIDE 1

Buffer sizing and Video QoE Measurements at Netflix

Bruce Spang, Brady Walsh, Te-Yuan Huang, Tom Rusnock, Joe Lawrence, Nick McKeown December 2, 2019

slide-2
SLIDE 2

How big should a buffer be?

Too big: packets wait for too long Too small: too many packets thrown away

slide-3
SLIDE 3

How big should this buffer be?

BDP: Villamizar and Song 1994 BDP/√n: Appenzeller, McKeown, Keslassy 2004 O(n): Dhamdhere, Jiang, Dovrolis 2005 O(1): Enachescu, Ganjali, Goel, McKeown, Roughgarden 2006

slide-4
SLIDE 4

Which is correct?

slide-5
SLIDE 5

It’s complicated

slide-6
SLIDE 6
  • 1. TCP New Reno (mostly) behaves

as expected

  • 2. Video performance varies
  • 3. Real routers complicate this story
slide-7
SLIDE 7

Our Experiment

slide-8
SLIDE 8

Catalog servers Uses spinning disks, cheaply stores entire catalog

slide-9
SLIDE 9

Offload servers Use SSDs to serve top ~30%

  • f content faster
slide-10
SLIDE 10

These three racks are called a stack

slide-11
SLIDE 11
slide-12
SLIDE 12

Make this buffer small… …and this

  • ne large
slide-13
SLIDE 13
  • 1. TCP New Reno (mostly) behaves

as expected

  • 2. Video performance varies
  • 3. Real routers complicate this story
slide-14
SLIDE 14

Large buffer has higher latency during congested hour

slide-15
SLIDE 15

Sometimes the large buffer has much higher latency

slide-16
SLIDE 16

Large buffer has lower loss during congested hour

slide-17
SLIDE 17
  • 1. TCP New Reno (mostly) behaves

as expected

  • 2. Video performance varies
  • 3. Real routers complicate this story
slide-18
SLIDE 18

Good buffer size: + Fewer rebuffers + Better video quality + Videos start faster Bad buffer size:

  • More rebuffers
  • Worse video quality
  • Videos start slower }

This happens when buffer is too large or too small.

slide-19
SLIDE 19

Site #2: A smaller buffer is better Reducing the buffer from 500MB to 25MB

  • 15.6% decrease in sessions with a rebuffer
  • 5.3% decrease in low quality video
  • 13.5% decrease in play delay
slide-20
SLIDE 20

Site #3: A smaller buffer is better Reducing the buffer from 500MB to 50MB

  • 22.1% decrease in sessions with a rebuffer
  • 7.0% decrease in low quality video
  • 14.8% decrease in play delay
slide-21
SLIDE 21

Site #1: A smaller buffer is worse Reducing the buffer from 500MB to 50MB +46.3% increase in sessions with a rebuffer +5.7% increase in low quality video

  • 5.9% decrease in play delay
slide-22
SLIDE 22
  • 1. TCP New Reno (mostly) behaves

as expected

  • 2. Video performance varies
  • 3. Real routers complicate this story
slide-23
SLIDE 23

Large buffer has higher latency during congested hour

slide-24
SLIDE 24

Remember how the large buffer has much higher latency…

slide-25
SLIDE 25

Servers have different very latency distributions

Min RTT (ms)

slide-26
SLIDE 26
slide-27
SLIDE 27

Server 1 Server 2 Buffer ISP …

What are we talking about?

slide-28
SLIDE 28

Line card #1 Line card #2 Line card #3 Line card #4

slide-29
SLIDE 29

VOQ #1 VOQ #2 VOQ #3 VOQ #4 VOQ #5 VOQ #6 VOQ #7 VOQ #8

slide-30
SLIDE 30 NETFLIX TEXT

Buffer architecture Server #1 Server #2 “Offload” VOQ “Catalog” VOQ ISP 100Gbps 2/3 1/3 Server #3

slide-31
SLIDE 31 NETFLIX TEXT

Traffic is fairly split when load is equal 40 Gbps 40 Gbps “Offload” VOQ “Catalog” VOQ ISP 100Gbps 67 Gbps 33 Gbps 40 Gbps

slide-32
SLIDE 32 NETFLIX TEXT

When one VOQ offers less than its fair share, it sees no congestion 50 Gbps 50 Gbps “Offload” VOQ “Catalog” VOQ ISP 100Gbps 90 Gbps 10 Gbps 10 Gbps

No delay!

slide-33
SLIDE 33

VOQs explain the RTT differences

Min RTT (ms) This VOQ is served faster This VOQ is served slower This VOQ is all

  • ver the place
slide-34
SLIDE 34

Switches prioritize long-tail content Same latency during uncongested hours Popular content is congested Long-tail content not congested

slide-35
SLIDE 35

Buffer Variable rate

Buffers are served at variable rates

slide-36
SLIDE 36
  • 1. TCP New Reno (mostly) behaves

as expected

  • 2. Video performance varies
  • 3. Real routers complicate this story
slide-37
SLIDE 37

Future steps

  • 1. Working with router manufacturer to explore ways of

setting buffer size

  • 2. What buffer size is best?
slide-38
SLIDE 38

Thanks!