Transcript Power Point

What did we learn at this workshop?
Dan Nessett, moderator
Usenix Special Workshop on Intelligence
At the Network Edge
San Francisco, CA
March 20, 2000
3/21/00
1
What did we learn at this workshop?
• What is the “edge” ?
- A host?
- An organizational or spatial boundary?
- A protocol layer boundary (e.g., above layer 3)?
• Edge intelligence can be used for :
- sub-multiplexing address/name spaces.
- scalable QoS provisioning.
- distribution of content processing and caching.
3/21/00
2
What did we learn at this workshop?
• Why put intelligence at the edge?
- improves scalability.
- raises difficult issue - do you make functionality transparent
(translucent?) or visible?
- when edge = end system, can shift responsibility for some
services [but will this really work?].
- allows offload of functionality from less capable devices to
“gateway” or “surrogate” with more capability.
- but, why use IP for this? Can’t less capable devices be
considered “peripherals”?
3/21/00
3
What did we learn at this workshop?
• Intelligence at the edge complicates (simplifies) management
- may make fault diagnoses/correction harder.
- may make management harder (heterogeneity).
- however, can improve fault tolerance.
- also, may make management more scalable (e.g., firewalls).
• How does end-to-end security affect edge intelligence?
- E2E encryption can hide data used for decision making.
- E2E encryption can prevent rewriting data in packets.
- Will affect web proxies and network device packet
classification significantly.
3/21/00
4
What did we learn at this workshop?
• Edge intelligence can break coherent architecture.
- May introduce more problems than it solves.
- Need an intelligent edge device “architecture” to help
minimize problems introduced into overall system.
• Need work on middle box discovery; auto-configuration.
3/21/00
5