lisp bof, ietf dublin, july, 2008 vince fuller (for the lisp crew) lisp+alt mapping system
TRANSCRIPT
LISP BOF, IETF Dublin, July, 2008
Vince Fuller (for the LISP crew)
LISP+ALT Mapping SystemLISP+ALT Mapping System
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 22
AgendaAgenda
• Mapping system design needs
• Ideas we considered• Brief summary of LISP+ALT• Open issues
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 33
LISP Internet DraftsLISP Internet Draftsdraft-farinacci-lisp-08.txtdraft-fuller-lisp-alt-02.txtdraft-lewis-lisp-interworking-01.txtdraft-farinacci-lisp-multicast-00.txtdraft-meyer-lisp-eid-block-01.txt
draft-mathy-lisp-dht-00.txtdraft-iannone-openlisp-implementation-01.txtdraft-brim-lisp-analysis-00.txt
draft-meyer-lisp-cons-04.txtdraft-lear-lisp-nerd-04.txtdraft-curran-lisp-emacs-00.txt
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 44
Mapping system: what and Mapping system: what and whywhy
• Need a scalable EID to Locator mapping lookup mechanism
• Network based solutions– Have query/reply latency– Can have packet loss characteristics– Or, have a full table like BGP does
• How does one design a scalable Mapping Service?
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 55
Scaling constraintsScaling constraints
• Build a large distributed mapping database service
• Scalability paramount to solution• How to scale:
(state * rate)• If both factors large, we have a problem
– state will be O(1010) hosts• Aggregate EIDs into EID-prefixes to reduce state
– rate must be small• Damp locator reachability status and locator-set changes
• Each mapping system design does it differently
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 66
Tough questions/issuesTough questions/issues• Where to store the mappings?• How to find the mappings?• Push model or pull model?• Full database or cache? Secondary storage?
• How to secure mapping entries?• How to secure control messages?• Protecting infrastructure from attacks• Control over packet loss and latency
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 77
LISP+ALT: What, How, LISP+ALT: What, How, WhyWhy
• Hybrid push/pull approach– ALT pushes aggregates - find ETRs for EID– ITR uses LISP to find RLOCs for specific EID
• Hierarchical EID assignment (geo?)– Aggregation of EID prefixes
• Tunnel-based overlay network• BGP used to advertise EIDs on overlay
– Use existing technology (and not DNS)
• Option for data-triggered Map-Replies
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 88
Legend:
EIDs -> Green
Locators -> Red
GRE Tunnel
Low Opex
Physical link
Data Packet
Map-Request
Map-Reply
ETR
ETR
ETR
ITR
ITR
EID-prefix
240.1.1.0/24
LAT
240.0.0.1 -> 240.1.1.1
1.1.
1.1
2.2.2.2
3.3.3.3
240.0.0.1 -> 240.1.1.1EID-prefix
240.0.0.0/24
1.1.1.1 -> 11.0.0.1240.0.0.1 -> 240.1.1.1
11.0.0.1 -> 1.1.1.1
ALT-rtr
ALT-rtr
ALT-rtr
ALT-rtr
ALT-rtr
ALT-rtr
12.0.0.1
11.0.0.1
?
240.0.0.1 -> 240.1.1.1
11.0.0.1 -> 240.1.1.1
? 240.0.0.1 -> 240.1.1.1
11.0.0.1 -> 240.1.1.1
?<- 2
40.1.1.0
/24
<- 240.1.2.0/24
< - 240.1.0.0/16
?
LISP+ALT in actionLISP+ALT in action
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 99
Issue: Data-Triggered Issue: Data-Triggered MappingsMappings
• ITR may forward data for “un-mapped” EID into ALT, attached to a Map-Request
• LISP Map-Reply returned from ETR to ITR, uses “native” path, installed in ITR cache
• ETR delivers attached data to end host• Subsequent traffic uses cached RLOCs• Scaling/complexity/performance issues• Is this (Data Probes) a good idea?
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1010
Issue: EID assignmentIssue: EID assignment
Provider A10.0.0.0/8
Provider B11.0.0.0/8
R1 R2
PI EID-prefix 240.1.0.0/16
10.0.0.1 11.0.0.1
ISP allocates 1 locator address per physical attachment point(follows network topology)
RIR allocates EID-prefixes(follows org/geo hierarchy)
SiteLegend:
EIDs -> Green
Locators -> Red
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1111
Separate EID/RLOC Separate EID/RLOC topologiestopologies
• ID/LOC separation avoids this dilemma
• EIDs uses organization/geo hierarchy• RLOCs follow network topology• Reduce global routing state through RLOC aggregation
• EID prefixes are not generally visible in global routing system
“Addressing can follow topology or topology can follow
addressing – choose one” –Y.R.
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1212
Issue: mapping system Issue: mapping system securitysecurity
• ALT can use existing/proposed BGP security mechanisms (SBGP, etc.)
• DOS-mitigation using well-known control plane rate-limiting techniques
• Nonce in LISP protocol exchange• More needed?
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1313
Issue: large-site ETR Issue: large-site ETR policypolicy
• ALT separates ETR discovery from the ITR-ETR mapping exchange– very coarse prefixes advertised globally
– more-specific info exchanged where needed
• Regional ETRs could return more- specific mappings for simple TE
• Alternative to current practice of advertising more-specific prefixes
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1414
Large-site ETR policy Large-site ETR policy exampleexample
• (someday, this will be a pretty, animated slide that shows how LISP and ALT can achieve the same “best exit” effect as advertising more-specifics with MEDs…today is not that day, unfortunately)
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1515
Issue: “low-opex” xTRIssue: “low-opex” xTR
• BGP configuration complexity is a barrier to site-multihoming
• Remove xTR/CPE BGP requirement:– ITR has “static default EID-prefix route” to “first hop” ALT router
– “first hop” ALT router has “static EID-prefix route” pointing to ETR
– originates EID prefix on behalf of ETR
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1616
Other issues to Other issues to considerconsider
• Who runs the ALT network?– What’s the business model?– Should it be rooted at/run by the RIRs?
– Different levels run by different orgs
– Should it be free?
• Others?
LISP BOFLISP BOF IETF Dublin, July, 2008IETF Dublin, July, 2008 Slide Slide 1717
Questions/Comments?Questions/Comments?
Slide Slide 1717
Thanks!
Contact us: [email protected]: http://www.lisp4.netOpenLISP: http://inl.info.ucl.ac.be