1. Carrier/Network Perspectives Cross-Layer Optimization (CLO) Bar Bof
© 2010 Telefónica Investigación y Desarrollo, S.A. Unipersonal

2. 01 Why cross layer? A little bit of history…
Layering was introduced to provide abstraction
Complexity for building applications is reduced
However… information is lost
Network layers are unaware of the detailed applications needs for the network
This can lead to an inefficient use of the network, and unexpected application behaviour
What can we do?
Include extra information in the headers (mark packets, extended info in RTP headers…)
Closer integration between application data and control (e.g. congestion notifications…)
Collaboration of network infrastructure (e.g. caching, resource) with application signaling, e.g. explicitly reserving bandwidth
Cross layer optimization

3. 02 Services Individual services
Services consumed by residential users. Typical services will include VoD and online gaming.
Individual services with high requirements.
Services consumed by companies / research centers involving a single application. Typical example will be a dedicated high bandwidth connection (backup service, connection to a remote device)
Aggregated services
The connection is used for a bunch of services with similar QoS requirements.

4. Understanding the application requirements03
Understanding the application requirements
Delivery parameters
Bandwidth demand (bit-rate) (peak demand)
Packet Loss: How sensitive is the service to packet loss?
Level of resiliency needed
Latency: Real-Time demand
Jitter: Latency variation
Video specific: Codec, Quality of the source material, e.g. quantization factor used in mpeg encoding
Elasticity: how well a service lends itself at having a different delivery rate compared to the play-out rate at the end-user. Real-time services or live services are not elastic.
Additional parameters, total volume: Is the service volume (in bytes) known? How small/large is this volume?
Symmetric: Is it symmetric or asymmetric (upstream-downstream) traffic. Are both directions equally important?

5. Understanding the application requirements04
Understanding the application requirements
Consumption parameters
Availability
Start-up delay
Consumption delay
Interactivity

6. Proposals and Solutions06
Proposals and Solutions
OGF Network Service Interface
Service interface between a requesting software agent and the provider software agent that offers and delivers a network service
SCAP
Implementation of transport with transparent intermediate buffering & deadline scheduling
Hop-by-hop request / reply mechanism
Meta-information in playmap

7. 05
Use case of service with high network interaction: Nomadic Virtual PC
The virtual PC: a virtual terminal which is executed on a remote host, and is accessed through the “remote desktop” capabilities of a standard Windows or MacOS terminal at home.
A nomadic Virtual PC service would require the transfer of data between servers in near real time (microseconds) timescale .
Example: consider a given user which is in the coverage area of a. The server at the local node asks the rest of the servers in the operator network for her virtual machine, which is immediately downloaded. As a result, the virtual PC is very close in propagation distance from the user, and the feeling of interactivity is very high.

8. Acknowledgements
Part of the ideas come from the research project RUBENS, and the ongoing projects MAINS

9. © 2010 Telefónica Investigación y Desarrollo, S.A. Unipersonal