5th GI/ITG KuVS Fachgespräch Next Generation Service Delivery Platforms

Notes of the 5th KuVS GI/ITG Workshop „NG Service Delivery Platforms“ with the topic „Advanced Service Delivery Platforms for Mobile Networks“

QoE optimization with network layer awareness on hybrid wireless networks (T. Melia, S. Randriamasy, Alcatel-Lucent, France; D. Munaretto, M. Zorzi, Univ. Padova/CFR, Italy)
– FP7 Medieval Project
– QoE based resource management in current BWA networks (LTE but also WiFi)
– Qoe Metrics for Video – Compare with reference and/or no reference
– Use the fact that different paths trough the networks means different QoE. Access/path selection based on QoE targets
– IETF ALTO (Application Layer Optimization) Working Group
— Describes attributes of caches and of their locations
— ALTO server provides ALTO clients with info on topology, routing cost, etc
– ALTO stores the network cost and needs to be combined with application layer metrics
– ALTO a tool for the operator to evaluate the performance of network caches
– In Medival ALTO not opened to users, ie operator does not reveal network-internals

Analysis of managed and OTT streaming services in mobile networks (J. Eisl, G. Kuhn, M. Lott, Nokia Siemens Networks, Germany; M. Varela, J. Prokkola, T. Mäki, J–P. Laulajainen, VTT, Finland)
Presented by M. Lott, Head of Service Control & Identity Management
– Compares „managed services“ (with QoS/QoE support by operators) and OTT services
– Different devices, different services (Mobile TV, Streaming Video, etc)
– Measurements via tcpdump, Wireshark, and tool developed by VTT
– OTT services mostly provided over HTTP, some use RTP/RTSP
– Microsoft Smoth Streaming and Apple HTTP Live Streaming are HTTP-based adaptive streaming protocols
– Conclusions: Rate adaptation not widely used, not supported by many servers, HTTP-based adaptations proprietary
– Conclusion: Managed services not significantly better than OTT services
— Paper author/NSN concludes: Means current QoS/QoE mechanisms are ineffective; more research required

FoG and Clouds: On Optimizing QoE for YouTube (T. Hoßfeld, F. Liers, T. Volkert, R. Schatz, Univ. Würzburg, Germany; Univ. Ilmenau, Germany; FTW Vienna, Austria
– Trend away from QoS towards QoE
– How to manage/optimize networks for QoE
– For HTTP-based video streaming there is no „low quality“, video will simply stall if too few resources
– Proposed: YouTube QoE metric that covers „stalling“
– Extensive study with human subjects to identify key criteria for „video quality“
– Mapping between „MOS and Number of Stalls“
– Conclusion: Users accept only very very few stalls (1 to no) and only very short stalls (exponential decrease of MOS)
– Equation for initial caching delay versus bandwidth provisioning

Application and Quality of Experience Aware Resource Management (D. Stähle, Univ. Würzburg, Germany)
– Look for ways to achieve win/win situation through network/user cooperation
– Possible with QoE as metric?
– And for Mesh-Networks?
– AquareYoumon tool by Uni Wuerzburg (Youtube QoE over mesh networks)
– YoMo – client-based application monitoring

MediaCloud – A Distributed Service Platform for Media Services (M. Bauer, S. Braun, P. Domschitz, Bell Labs, Alcatel-Lucent, Germany)
– Pardigm shift expected away from „Big Iron“ approach (datacenter centric) towards distributed „MediaCloud“
– A new approach to „building distributed services“
— Flow-driven model, Atomic execution model;
— Work on media chunks, not IP packets
– G-Streamer framework goes a bit into this direction

Seamless Service Provision in P2P Service Overlays (K. Panitzek, I. Schweizer, M. Ikramy and Max Mühlhäuser, TU Darmstadt, Germany)
– P2P service overlays – distribution and execution of applications, composition of services
– Seamless service migration between peers
– MudoCore middleware – for code migration between peers
– Chord DHT – for distributed service registry

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert

Diese Website verwendet Akismet, um Spam zu reduzieren. Erfahre mehr darüber, wie deine Kommentardaten verarbeitet werden.