Large-scale Media Processing on Cloud - Cloud Asia 2016 PANEL DISCUSSION

Download Large-scale Media Processing on Cloud - Cloud Asia 2016 PANEL DISCUSSION

Post on 08-Feb-2017

94 views

Category:

Software

0 download

Embed Size (px)

TRANSCRIPT

  • +

    Large-scale Media Processing on Cloud

    (Discussion Panel - Introduction)

    Dr BHOJAN ANAND (Panel Moderator)

  • +Let's Make this Interactive

  • +Large-Scale Media Processing in Cloud

    Storage

    Search

    Retrieve

    Encrypt

    quality control

    Transcode

    Analytics

    DRM

    File Transfer Stream

    Ever increasing Media Data: 400 hours of video uploaded/min & 3 trillion views/year in Youtube

    Compress

  • +Large-Scale Media Processing in Cloud

    Consumer Demand (choices

    and control)

    Time pressures

    (more OTT providers)

    Cost Pressures

    (resources)

    Eg. Media Entertainment Industry

    scalability, agility and

    cost flexibility

    and

    Calling for

    Cloud

  • +Discussions .

    PANEL MEMBERS INTRODUCTION PANEL MEMBERS PRESENTATION DISCUSSIONS Q&A

  • +

    CLOUD GAMES

    - Multiplayer Cloud Games

    - Mobile or Resource limited devices

  • + Multiplayer Games

    client

  • +Multiplayer Games

    nMobile Games and Game Maps

    nCharacteristics (FPS games) nHighly interactive, Closed maps with occlusions, 40-80

    fps

    n~20 pps server to client (in

    regular interval, cannot burst!)

    n~40 pps client to server (in regular interval, cannot burst!)

    nLatency more than 180ms makes the game play annoying and not playable

  • +Multiplayer Games

    nCharacteristics (MMORPG games) nSlow speed games, with few high interactive areas nHuge (100s of square miles) open maps with few

    occlusions. [Estimated time taken to run from one end to another].

    n100s to few 1000s of players per server

    n40-60 fps n~15 pps server to client n~20 pps client to server nLatency more than 400ms

    makes the game play annoying and not playable

  • +Multiplayer Games

    nClient side Latency Hiding Techniques

    Immediate Response Prediction

    Local Perception

    Filters

    Require Intelligent

    Clients!

  • + Mobile Multiplayer Cloud Games

    No download, no patches, no disks, no installations

    CPU/GPU load is reduced to save energy

    Centralized control. Hard to cheat

    Scalable

    Portable

  • +Mobile Multiplayer Cloud Games Key Challenges

    - All these operations should happen serially in few ms!

    - Requires

    high sever bandwidth for Video Streaming

    Inte

    ract

    ion

    Del

    ay

  • +Multiplayer Cloud Games

    nCould Games need to handle additional latencies nCloud Processing (Cloud Overhead) nEncoding Video nCompression nWorse in Mobile Environments due to prevailing

    Network Latency and Jitter nEven with LTE it takes 70ms to closest data

    center, Goes upto 200ms

    Key Challenges

    Conventional Client Side Latency Hiding Techniques are not Possible! Client is Dump!

  • +Multiplayer Cloud Games

    nthe company had deployed thousands of servers that were sitting unused, and only ever had 1,600 concurrent users of the service worldwide.

    http://www.theverge.com/2012/8/28/3274739/onlive-report

    Active cloud gaming providers GameNow, GameFly Streaming, G-cluster, Gface, Kalydo, LoudPlay, Leap Computing, Liquidsky, PlayGiga, PlayKey, PlayStation Now, NVIDIA GeForce Now, Turbo Games

  • + Possible Approaches: Gamelet System (case for micro-cloud!)

    nGamelet is a minimal-set hardware required to run, render and stream 3D games placed in the same local network or few hops (most of the time one or two hops) away from the mobile client.

    nUses distributed rendering to counter the limitations of renderer HW resources

    Render at next hop!

    Bhojan Anand, Aw Jia Hao Edwin , "Gamelets - Multiplayer Mobile Games with Distributed Micro-Clouds, IEEE-ICMU 2014

  • +Demo

    n Look for Gamelets - Multiplayer Mobile Games with Distributed Micro-Clouds in Youtube

  • +Speculative Rendering in Cloud

    nSpeculation/Prediction for Navigation and Camera Movement

    nImpulse Events - Explore multiple outcomes in parallel and render for all possibilities

    Outatime: Using Speculation to Enable Low-Latency Continuous Interaction for Mobile Cloud Gaming, Kyungmin Lee, David Chu, Eduardo Cuervo, Johannes Kopf, Yury Degtyarev, Sergey Grizan, Alec Wolman, Jason Flinn, Univeristy of Michigan, Microsoft Research, Mobisys 2015

  • +

    THANK YOU Questions?

Recommended

View more >