Transcript The Tapestry Solution

Tapestry
Software Infrastructure Enabling
Virtual Private Networks
K. Balakrishnan, N. Chokshi, R. Huang, A.
Konrad, S. Kumar, O. Sakdamnuson, B. Zhao
Presentation Outline






The Pitch
The technology: Tapestry
Three Potential Applications
Introduction to Virtual Private Networks
Solving VPN Pains with Tapestry
Strategy
The Elevator Pitch






For enterprises and homes
Who need mobile, secure, and reliable data
communication
The Tapestry-enabled VPN is a softwarebased VPN solution
That delivers a scalable, reliable solution
Unlike existing VPN solutions
Our product provides portability, lower cost
QoS and self-management capabilities
Technology Motivation

Networks growing exponentially



Huge potential for applications


Size, speed, reach
Issues of frequent faults, bottlenecks, and
complex manageability issues
(IF we can solve these issues)
Solution: a network layer that is



Scalable over the wide-area
Tolerant of multiple failures
Self-monitoring and optimizing
Tapestry


PhD Thesis work: Ben Zhao, EECS, UCB
Wide-area network infrastructure





Software components overlaid on IP
Find nearest objects given ID
Route messages to nodes and objects efficiently
around faults
Intelligent software communicating with each
other to collectively monitor network
Performance with minimal overhead
Key Novelty

Distributed algorithms & data structures






Provide routing from A to B
Each hop gets close to B in name
E.g.: node 3459 going to node 9621
3459  3251  1521  0621  9621
Numerous paths between any 2 pts
Each node keeps data proportional to Log of
network size
Efficient:

Minimal overhead in storage, routing distance, and
bandwidth
So what…?

Wide-area application infrastructure



Easily build wide-area applications
Transparently inherent Tapestry properties
Research applications:


Bayeux: Wide-area scalable multicast
Silverback: Global-scale archival file
storage
Three Potential Applications

Streaming Multimedia


FreeFlow (Akamai), FastForward (Inktomi)
What we add: efficient multicast over wide-area


Electronic Automatic Content Distribution



Victoria’s Secret Fashion Show
Marimba, Pointcast, BackWeb
What we add: efficient content push to consumers
Virtual Private Networks


Nortel, Checkpoint Software
What we add: low cost, software-based QoS
Most Compelling: VPNs
Application
Attractiveness of Rationale
Underlying
Market
Multimedia
Streaming
Low
400 M (2000)
Uncertain future demand
EACD
Low
500 M (2000)
Flat growth
Unproven market
VPNs
High
2B annual (2000), 3.5B (2001)*
High projected growth
(home VPNs)
* Data from IDC
What is a VPN?


Virtual private networks provides reliable,
secure and mobile data communications
Key benefits



Network managers: increases the reach of the
corporate network cost efficiently
Remote users: secure access to corporate network
Corporations: secure communications via
extranets to business partners
Virtual Private Networks
Current VPN Pains



High cost of dedicated VPN access
provider
High cost of management /
maintenance
Quality of Service is Expensive


PPNs have QoS, at high cost
“Cloud”-based VPNs provide no QoS
The Tapestry Solution

Flexible access



Self management and maintenance



Active fault-detection and repair
Load-balancing, resilience to DoS attacks
Software-based Quality of Service


Small, portable software component
Transparent to choice of location, ISP/network
QoS guarantees without dedicated lines
Status: working simulations; finishing prototype
Snapshot Result
TCP Connection Packet Loss
Standard TCP/IP
Tapestry-enhanced TCP/IP
50%
Packet Loss (%)
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
650
850
1050
1250
Background Network Traffic
1450
Brief Summary

Identified 3 potential Tapestry applications




VPN application is most compelling




Multimedia Streaming
Electronic Content Distribution
Virtual Private Networks (VPNs)
Larger market opportunity
Est. $2bil in 2000, $3.5bil in 2001
Tapestry can solve pains for VPNs
So, how can we bring Tapestry to market?
Strategy
License the
technology to VPNs
Tapestry
Develop own VPN
product &
incorporate Tapestry
Strategy Evaluation
Strategy
Customers
Competition
Licensing
VPN Service Providers MIT, AT&T, Microsoft
(Nortel, MCIResearch
Worldcom, AT&T)
Do not address QoS
Weaker performance
guarantees
VPN Product
Enterprises
ISPs
Considerations
Lesser relative risk
Limited revenue
opportunities
VPN Service Providers Longer time to
(Nortel, MCImarket
Worldcom, AT&T)
More established
competition
Market Segment (Product)

Initial



Expanding




Business type: small to medium (SMEs)
Industry: manufacturing, education, government
Business type: large
Industry: healthcare, financial services, high-tech
Application: remote access or extranet access
New, value-added services


Network design and Consulting
Global scale applications
Deployment and Pricing

Deployment Issues




Critical mass of servers necessary
Initial partnerships w/ Akamai/Inktomi
Leverage self-propagation effect
Pricing Model


Initial fee for deployment
Cost per seat