Distributed Peer-to-peer Name Resolution

1
Distributed Peer-to-peer Name Resolution

• Christian Huitema
• Architect, Windows Networking
• Microsoft Corporation

2
Vision The user to user web!

• Enable each PC to be a server
• Serve ad-hoc groups
• Promote and use IPv6, directly attach the users
  to the Net
• Provide a server-less DNS for IPv6
• Platform for developers

Mainframe
PC
Web
P2P
Centralized
Personal
3
Making Windows a great platform for P2P
Firewall Traversal
Sockets, DNS, files, print, .
Firewall, NAT
IPv6
Multicast?
Security?
DNS?
Graphs
Group Membership
P2P Names
4
Naming the key to P2P development

• Provide each computer with a name
• Each computer is a server
• Names must be stable
• Resolve names to addresses
• Also name groups, users
• Find a member of a group,
• Find the location of a user

5
Peer-to-Peer Name Resolution

• Identity 128-bit number
• No trademark!
• Nodes manage a cache of records
• Identity
• IPv6 address
• Proof (certificate)
• Secure, Scalable, Robust, Private, Usable

Look up initiated
6
Design Goal 1, Security

• Names should resolve to the intended address,
• One should not be able to fake an identifier
• But one should not impose a central server

Hello, I am alice_at_example.com
7
Secure use certificates, prove the name

• Record Identity, IPv6 Address, Proof
• Sign identity, address with local private key
• Check signature before placing in cache
• Host Identity hash (public key)
• Group member Id hash (group key)
• Group membership, certified by group key
• Generic hash (authoritys key, Name)
• Validity certified by authority

8
Design Goal 2 Scalable
1,000

• Scale to the size of the Internet!
• Assume at least one name per node.
• Scale without a central server.

10,000
100,000
1,000,000
10,000,000
100,000,000
1,000,000,000
9
Scalable a hierarchical cache

• Circular space, centered on self
• Cache levels
• First level 360
• Second level 36
• Etc.
• Query
• Forward to nearest
• Gain one level
• log base X of (N) levels, 2.X Entries per level

10
Scalable cache management

• Initial cache
• from memory, from neighbor
• Proactive requests, tell the network
• Update
• Learn from queries, requests
• Try to fill each level, then replace
• Finest level
• Learn about all immediate neighbors

11
Design Goal 3 Robust

• No central point of control, no central point of
  failure.
• Large distributed system some nodes will fail at
  any time.
• Uncontrolled systems some nodes may be actively
  conspiring against us but hopefully not too
  many.

12
Robust reputation randomization

• Neighbor reputation from query completion
  verification
• Forward query random choice, use reputation
• Cache replacement random reputation
• Timers, removal of failing entries from the cache

13
Design Goal 4 Privacy

• Some information is sensitive location,
  membership in groups, documents
• Control who can learn what you publish!

14
Privacy control the scope of publication

• Peer-to-peer name cloud set of caches
• Parallel clouds
• Ad hoc, on the same link
• In the same site
• Within a group (group authority)
• The whole Internet
• Enforcement
• Check membership of requestor
• Dont forward outside of scope
• Dont cache out of scope records

15
Design Goal 5 immediately usable

• On the Internet, Name Domain Name
• Used for Web Page, E-Mail, FTP
• Limited availability, requires a contract and a
  permanent IP address.

16
Usable gateway to the DNS

• Global DNS name
• ltIDgt.pnrp.net
• Use in URL, e-mail
• Enables bootstrap
• Central DNS gateway
• Publish NS record,
• Join P2P cloud,
• Forward requests
• Local gateway
• P2P aware resolver

Target Node Reply on the same path








Look up initiated

PNRP.net
17
Peer-to-peer Naming Summary
Target Node Reply on the same path

• Numbers are our friends hosts, groups, URL
• Scale to the size of the Internet, peer-to-peer,
  no server.
• Prototype tested with 50,000 nodes
• Secure, Private, Robust, Usable

Look up initiated

PNRP.net
18
Peer-to-Peer Roadmap

• UPnP Enabled NATs End of 01
• IPv6 Service Deployment End of 01
• P2P SDK Design Review End of 01
• P2P SDK beta Q1-02
• P2P SDK v1 Q3-02

19
Call to Action

• Support IPv6 in your Applications
• Leverage UPnP NAT Traversal Today
• Participate in Software Design Review