Using Prices to Allocate Resources at Access Points

Using Prices to Allocate
Resources at Access Points

Jimmy Shih, Randy Katz, Anthony Joseph
Network
Resources
Access
Point A
Access
Point B
One
Administrative
Domain
User A
User B
Problem Statement


We can solve congestion at access points by adding
capacity, performing admission control, balancing
loads between the access points, allowing advance
reservation or constraining usage.
We plan to explore:
 How effective prices can affect users’ behaviors
when used for admission control and load
balancing at the access points.
 How to design resource allocation schemes to
minimize the adverse effects of pricing. E.g
instability, overhead.
State of the Art on
Congestion Pricing

Prior theoretical & simulation work suggest
that congestion pricing can



Allocate resources to those who value them the
most.
Provide users with incentives to adjust their
behaviors instead of constraining them.
However, congestion pricing can also


Cause users to experience unpredictable cost and
performance.
Require system overhead like accounting.
Key Concepts




Perform congestion pricing at the application
layer => Only need to provide users with a
web interface containing the current prices.
Perform congestion pricing at the access
points => Only need to modify the access
points.
Use congestion pricing for admission control
when users are willing to wait for resources
or adjust their usage while using them.
Use congestion pricing for load balancing
when some access points are clearly better
than others.
Using Prices to Allocate
Resources at a H.323 Gateway

Use prices for admission
control and load balancing on
connection oriented streams.


Use prices to allocate the
bottleneck resource, the
number of phone lines at the
gateway.
Use prices to encourage
users to use their computers
instead of telephones to
minimize using the
bottleneck resource.
Internet
H.323 PSTN
Gateway
Possible User Web Interface
Current Price for Using
Your Computer:
10
Tokens/min
Next Minute Price for
Using Your Computer:
20 Tokens/min
Current Price for Using
Your Telephone: 15
Tokens/min
Next Minute Price for
Using Your Telephone:
35 Tokens/min
Packet Loss Rate When Using Your Computer:
3%
Handoff the Current Call to Your Telephone:
(510) 642-8919 Yes?
Handoff the Current Call to Your Computer:
center.cs.berkeley.edu Yes?
Using Prices to Allocate
Bandwidth at an Access Router

Use prices for admission control on
connectionless resources.


Use prices to decide whose
incoming and outgoing packets to
drop during congestion.
Inform users the amount of
bandwidth they are using and
allow them to reserve the amount
they need.
Possible User Web Interface
Current Price Per
Mbyte Per Sec:
10 Tokens
Tokens Left for
Today:
200
Tokens
Current Usage: 4 Mbyte Per Sec
How Much Bandwidth to Purchase:
3 Mbyte Per Sec
Users in a
Local Area
Network
Access
Router
Internet
Most Willing to Pay Per Mbyte Per Sec:
30 Tokens
Accomplishments



Design the appropriate policies for using
congestion pricing at a H.323 gateway and at
an access router.
Fall of 1999 - Alpha testing of using
congestion pricing at a H.323 gateway with
users in our research group.
Spring of 2000 – Beta testing of using
congestion pricing at a H.323 gateway with
50 users in the EECS department.
Plans for Success


Simulations:
 Estimate the potential benefits and harms of using
congestion pricing versus a first come first serve policy.
 Simulate when users are not sensitive to changing prices or
require long times to react to changing prices.
 Use calling patterns from the H.323 gateway’s Beta testing.
 Use IP traces from the Internet Traffic Archive.
Deployments:
 In August, provide students in the UC Berkeley dorms with a
H.323 gateway service that uses congestion pricing.
 In the fall, replace our research group’s subnet router with a
Nortel Java Programmable router that performs congestion
pricing.
Summary and Conclusion



See if congestion pricing really works in practice.
Explore the right policies for using congestion
pricing. E.g. how to ensure price & system
stability, minimize starvation, prevent denial of
service attacks, etc.
Explore the economic issues of using prices. E.g.
amount of incentives required for users to choose
congestion pricing over flat rate pricing, to use
their computers instead of telephones, etc.