Rethinking Traffic Management: Using Optimization Decomposition to Derive New Architectures

1
Rethinking Traffic ManagementUsing Optimization
Decomposition to Derive New Architectures

• Jennifer Rexford
• Princeton University
• Jiayue He, Maayan Bresler, and Mung Chiang

Or, how I learned to stop worrying and love
optimization theory
Or, why I walked across the street to the EE
department
2
Traffic Management Today
Operators Traffic Engineering
Link weights
Source rates
INTERACTION IGNORED
Routing
Users Congestion Control
Routers Shortest Path Routing
Division of labor evolved organically
3
Top-down Redesign
Problem Formulation
Optimization decomposition
Distributed Solutions
Compare using simulations
Heuristic
Touch of human intuition
Final Protocol
Using optimization to design protocols
4
So, What Do We Want?

• Users maximize utility, fairness

• Operators load balancing, robustness

minimize sum(f(ul))
maximize sum(Ui(xi))
5
Balanced Objective
Congestion Control User Performance
Traffic Engineering Network Robustness
Can be at odds!

• max. ?iUi(xi) - w?lf(ul)

Weight of penalty
Tweaking w allows for striking the balance
6
Decomposing the Optimization Problem

• Multiple paths with flexible splitting
• Tune sending rate on each path
• Based on feedback from the links

z11
z21
z31
Provably optimal and stable!
7
Several Different Decompositions

• Multiple decomposition techniques
• All multi-path protocols with link feedback
• Different link feedback and rate updates
• Different number of tunable parameters
• Math doesnt answer all questions
• Sensitivity of tunable parameters
• Speed of convergence
• Need for simulations and intuition
• Combining the best of each algorithm

8
New Division of Functionality

• Sources end hosts or edge routers?
• Feedback implicit or explicit?
• Computation centralized or distributed?

Mathematics leaves open architecture questions
9
Experimental Requirements

• Revisiting traffic management
• Routers multi-path routing/forwarding
• Hosts end-host rate adaptation
• Realistic traffic patterns
• Realistic synthetic models
• Real users
• Composition with others
• Beyond distributed algorithms
• To a real network architecture