Multicast Applications: Challenges & Solutions

Download Report

Transcript Multicast Applications: Challenges & Solutions

IP Multicast Applications:
Challenges & Solutions
Bob Quinn
Stardust Forums, Inc.
IP Multicast Initiative
1
Agenda
•
•
•
•
ID Motivation & Purpose
ID Scope
Multicast Application Taxonomy
Primary Requirement Challenges
– Heterogeneous Receivers
– Reliable Delivery
– Secure Multicast
2
Motivation for the Internet Draft
IP Multicast Initiative - ipmulticast.com
– What: Multi-vendor Forum
• Members are International and Cross-Industry
– Why: Promote adoption of IP Multicast
• Overcome the Chicken-and-Egg problem
– How: Education
• Evangelize its benefits, and describe its details
• Publish documentation and Host events
• Informational Internet Draft...
3
Purpose of the Internet Draft
• Informational
– <draft-quinn-multicast-apps-00.txt>
• A Roadmap for Application Developers
– An Orientation Tool
– “You are here...” in terms of
• Application Type
• Application Requirements
– “Here’s where to go now...”
4
Ultimate Goal of Internet Draft
• Help Application Developers to Avoid:
– Reinventing Wheels
– Spinning Wheels
– Running Roughshod over IP Networks
• encourage “network-friendly” application designs
• “heads-up” when “you can’t get there from here”
5
Scope of Internet Draft
• Focus is on Applications
– not Infrastructure or Mechanics
• Assumption: Multicast-Enabled Network:
– Already Exists
– Transparent to Applications
• Assumption: Developer aware of mechanics
– Refer to Maufer/Semeria ID for details
<draft-ietf-mboned-intro-multicast-03.txt>
6
Multicast Application Taxonomy
• Sender/Receiver Relationships
– Differentiate Multicast from Unicast
– Characterize all Multicast Applications
• All Multicast Apps are one of Three Types:
– One-to-Many
– Many-to-Many
– Many-to-One
7
One-to-Many Applications
• Analogous to the TV/Radio Broadcast Model
– Useful and valid analogy, but not all there is...
• Push Media: headlines, weather, sports...
• File Distribution and Caching
• Announcements: network time, session schedules,
randoms, keys...
• Monitoring: stock prices, sensors, security,
manufacturing...
8
Many-to-Many Applications
•
•
•
•
•
•
•
•
Multimedia Conferencing: A/V and whiteboard
Synchronized Resources: Database updates...
Concurrent Processing: Distributed & Parallel
Shared Editing and Collaboration
Interactive Distance Learning
Distributed Interactive Simulations (DIS)
Multi-Player Gaming, Chat Groups
Jam Sessions
9
Many-to-One Applications
•
•
•
•
•
•
Many are request/response...
Resource Discovery: Service Location, Anycast
Data Collection
Auctions
Polling
Juke Box
10
Multicast App Requirements
• Bandwidth, Delay and Jitter
– Multicast Apps are no different from Unicast
– Useful to characterize needs nonetheless...
• Multicast is a Special Case when Servicing:
– Heterogeneous Receivers
– Reliable Data Delivery
– Security
11
Heterogeneous Receivers
• Adapting to receiver rates, delays and jitter
– Difficult with unicast, but worse with multicast
• Strategies and Challenges:
–
–
–
–
Feedback loops - Danger of implosions
Forward Error Correction - Uses bandwidth
Shared Learning - Privacy Issues
Local Recovery - Management Issues
• No standards yet. Research effort informal.
12
Reliable Data Delivery
• Receivers tolerant of delay, but not loss
– TCP model doesn’t translate (implosion)
• Strategies and Challenges:
– NAKs - implosion potential still exists
– FEC, Shared Learning, and Local Recovery
are all also relevant
• No standards yet.
• RMRG in IRTF is defining the problem
13
Multicast Security
• Receivers and Senders are wary
– Don’t trust data, or each other, or anyone else
– Unicast is hard, and multicast is HARDER
•
•
•
•
Challenges: authentication and key distribution
Strategies: many and varied
No standards yet.
SMuG in IRTF is defining the problem
14
Observations
• Three Primary Requirements share many
of the same Strategies & Challenges
– Common issue is dealing with asymmetry in the
sender/receiver relationship
– e.g., Local Recovery - smarts in the net
– Solutions may have cascade effect (?)
• Prevailing Consensus: we need more than one
protocol to satisfy the varying application
requirements in each category
15
Evaluation
• Is this of value?
– Should this be a draft from MBoneD WG??
• Any Questions? Comments? Suggestions?
– Send to <[email protected]> & mboned mail-list
• Also see http://www.ipmulticast.com
16