Topics in Networking
· Sources of increased traffic (local, long-haul)
· Directions in communications protocols
· Problems of scale in large internetworks
· Voice in local area packet-switched networks
Sources of Increased Traffic (Local)
· 13 Year phenomenon
Laser printing
High-performance workstations
Packet-switched wide-area distributed communications?
· Home (small business/institution) computing
Chicken and egg -- substantial market at acceptable rates, none at today's
New applications areas
Diskless workstations
Remote window systems (e.g., Sun's NEWS)
Remote publication-quality filing and printing
Bandwidth requirements
(Response-time, not throughput)
9600 bps marginal
56 kbps tolerable
ISDN, packet, Cable?
CSDC (near-miss, so far)
Ideal: packet-switched service at T1 data rates
· Support of remote connections without bypass
Fiber plant enhances desirability
Switched service
Sources of Increased Traffic (Remote)
· 56K backbone now in Xerox
· Very large latent market
Critical mass (10,000 terminals in Xerox network)
Function of cost and performance
· Video conferencing beginning to take off
· Portland experiment
Will this sort of thing catch on?
How soon?
· Next increment: image transmission within integrated publication/telecopier/office systems
Protocols
· Interoperability a major issue
X.400 (electronic mai)
SNA
TCP/IP (DARPA protocols)
ISO (when?)
ISDN? (not a complete architecture)
· ISO 7-layer model versus simpler models:
Link and transport layers intact
Service layer:
Remote procedure call (RPC) abstraction; lightweight (client-invisible) session layer
Bulk data transfer, controlled by RPC
Higher layers more properly treated as distributed applications, largely independent of communications
· New DARPA/NSF high-performance protocol initiatives
Problems of Scale
· Can't control load through denial of service in a datagram-oriented network
· Naming and resource management require distributed databases
· Update and query problems in distributed databases unsolved, in general
User-perceived performance problems
Algorithms that scale catastrophically
· Relative immaturity of network measurement/management tools
· Recent case study within Xerox internet
Distributed name database designed for 30 networks, servers
network contains 300 networks, servers
n3 update algorithms recently detected, replaced
· Other exciting opportunities for research based on large, real, malleable internetwork
Congestion control, flexible routing, . . .
Voice in Local Area Packet-Switched Networks
· Technically feasible
Primary obstacle is packet delay
CSMA/CD (Ethernet) issues not the major ones
Requires proper internetwork engineering
50-phone prototype system in daily use at PARC
BellCore work has reached similar conclusions
· Technically desirable
Integration of voice/data transmission
Integration of voice recording, synthesis, recognition services
Enhanced control, broadcast services, etc.
· Economics not yet clear
· Primary challenges are systems challences
Continued reliability of base telephone features
Workstation control of voice facilities
Need for overall voice systems architecture