Network Localized Mobility Management using DHCP
description
Transcript of Network Localized Mobility Management using DHCP
Network Localized Mobility Management using DHCP
NETLMM Problem Space
global mobility(out-of-scope)
Global Internet
NETLMMDomain A
NETLMMDomain A
NETLMMDomain B
MobileNodes
localizedmobility
NETLMM Goals
• support NETLMM domains as small as a home network or as large a major operator network, e.g., metropolitan region WiFi
• MNs keep same addresses/prefixes as they move within a NETLMM domain (global mobility out-of-scope)
• support session continuity across mobility events• avoid routing churn by having Mobility Anchor
Points that aggregate the NETLMM domain (as opposed to tracking node mobility via a routing protocol)
NETLMM Domain
Backhaul Network
Access Networks
Mobile Nodes (MNs)
Mobility AnchorPoint(s) (MAPs)
Access Rtrs (ARs)
NETLMM Using DHCP
• Let each MN be a DHCP client
• Let each AR be a DHCP Relay
• Let each MAP be associated with a DHCP server (no need for them to be co-located)
Model of Operation
• MN discovers ARs via RFC2461 Router Advertisements (RAs)
• If RAs contain prefix options, MN can configure addresses using RFC2462, then “register” them with the network by sending DHCP Solicit/Request with IP address options
• If RAs contain no prefix options, or if prefix delegation is desired, MN requests prefixes by sending DHCP Solicit/Request per RFC3633
• AR relays DHCP Solicit/Request to a DHCP server associated with a MAP
Model of Operation (cont’d)
• DHCP server registers addresses/prefixes, then issues “create tunnel”; “route add” to update MAP IP forwarding table(s)
• DHCP server sends reply to MN which is intercepted by AR; AR performs a local “route add”
• Now, traffic from the Internet destined to MN flows through the MAP(s) and is directed to the correct AR
• If MN moves to a new AR, MN issues a DHCP Confirm which causes the MAPs and ARs to update their IP forwarding tables
Route/Tunnel Configuration after MN config’s address/prefix via AR1
MN1
AR1
MN1AR1AR1tunnel
MN1on-link
AR2
Route/Tunnel Configuration After MN moves to AR2
MN1
AR1
MN1AR2AR2tunnel
MN1on-linkAR2
Additional Considerations
• Works with IPv4 as well as IPv6 (IPv6 has some advantages)
• Supports DHCPv6 prefix delegation (delegated prefixes move along with the MN)
• tunnels from MAPs to ARs can be unidirectional• Explicit messaging between MAPs and ARs
might be better than implicit route add/delete based on DHCP messages – being worked in IETF NETLMM wg
Additional Considerations (cont’d)
• With multiple ARs on the link, ambiguous as to which AR is selected in MAP forwarding tables – MN can assert AR selection by sending L3 multicast DHCP Solicit/Request to unicast L2 address of a specific AR
• global addressing goes through MAPs, but efficient local communications can be supported using IPv6 ULAs (could result in dropped calls)
• Since MNs can move freely between access networks, Redirects could cause dropped calls. ARs on NETLMM links should therefore not send redirects.
Issues
• can DHCP Confirm be used to test whether a delegated prefix is appropriate for the new link. If not, why not?
• with all global addresses/prefixes delegated by DHCP server, no need for DAD on NETLMM links?
• link-local addresses can also be registered with DHCP server. Again, no need for DAD?
MANET Autoconfiguration using DHCP
MANET Autoconf Problem Space
Provide Network(s)
MANETs
MANET Gateways(MGs)
MANET Routers
MANET Routing Alternatives
• MANET routing as a L2 mechanism w/no L2 multicast flooding – MANET looks like an NBMA link that connects routers/gateways (no gateway discovery; not considered further)
• MANET routing as a L2 mechanism w/L2 multicast flooding - MANET looks like a bridged campus LAN (no special MANET Autoconf extensions needed)
• MANET routing as a L3 mechanism w/no L3 multicast flooding – MANET looks like multiple links w/no multicast (no gateway discovery; not considered further)
• MANET routing as a L3 mechanism w/L3 multicast flooding – MANET looks like multiple links w/MANET-wide (site-scoped) multicast (subject of this proposal)
MANET Autoconf Goals
• MANET Routers (MRs) must be able to discover MANET Gateways (MGs) even if they are multiple L3 hops away
• MRs must be able to obtain global IP address/prefix delegations
• support for multiple MGs• support for intra- and inter-MANET
mobility for MRs• Support for both IPv6 and IPv4
MANET Autoconf Using DHCP
• Let each MR and MG configure a MANET Local Address (MLA) for routing protocol operation; local communications (for IPv6, likely to be RFC4193 ULAs)
• Let each MR be a DHCP client• Let each MG be a DHCP Relay/Server• Let there be a means for MRs and MGs to
send “Extended” RS, RA and DHCP messages
Extended RS, RA and DHCP Msgs
• normal RS/RA/DHCP message configured per RFC2461, RFC3315, RFC3633
• message encapsulated in outer IP header with:– src = MLA of sender– dst = Site-scoped multicast, or MLA of target– Hop Limit = small integer value (e.g., 2, 5, 10)
• message “tunneled” to one or more recipients
Extended RS, RA and DHCP Msgs
NormalRS/RA/DHCP
Message
src=LL/Unspecdst=All_*_Multicast
Hop Limit = 255
src=MLA(sender)dst=Site-scope Multicast
or MLA(target)Hop Limit = 2,5,10,etc
Message
Inner IP header
Outer IP header
Model of Operation
• MR discovers MGs via Extended RAs (ERAs) (MR sends ERSs if necessary)
• If ERAs contain prefix options, MR can configure addresses using RFC2462, then “register” them with the network by sending Extended DHCP Solicit/Request with IP address options
• If ERAs contain no prefix options, or if prefix delegation is desired, MR requests prefixes by sending Extended DHCP Solicit/Request per RFC3633
• MG decapsulates the Extended DHCP Solicit/Request and relays it to a local DHCP server or a server in the provider network
Model of Operation (cont’d)
• DHCP server sends reply to MR which is intercepted by MG; MG performs a “route add” and “create tunnel” for MR
• MR receives the DHCP reply and performs a “route add” and “create tunnel” for MG
• Now, packets from the Internet destined to MR are directed to MG which tunnels them to MR, and packets from MR destined to the Internet are tunneled to MG
• If MR moves to new MG, it sends an Extended DHCP Confirm which causes MGs to update their IP forwarding tables
Route/Tunnel Configuration after MR1 conf’s address/prefix via MG1
MR1MLA(MR1)MLA(MR1)tunnel
MG1
MR1defaultMLA(MG1)MLA(MG1)tunnel
MG2
Route/Tunnel Configuration after MR1 moves within MANET
MR1MLA(MR1)MLA(MR1)tunnel
MG1
MR1defaultMLA(MG1)MLA(MG1)tunnel
MG2
Route/Tunnel Configuration after move to MG2 in same MANET
MR1MLA(MR1)MLA(MR1)tunnel
MG1
MR1defaultMLA(MG2)MLA(MG2)tunnel
MG2
Route/Tunnel Configuration after move to MG3 in new MANET
MG1
MG2
MR1MLA(MR1)MLA(MR1)tunnel
MR1defaultMLA(MG3)MLA(MG3)tunnel
MG3
MG2, MG3 connectedto same provider
Additional Considerations
• Compatible with “NETLMM using DHCP”• Works with IPv4 as well as IPv6 (IPv6 has some
advantages)• For IPv4, need a new option (“MLA Option”) to
inform relay of MR’s MLA for last-hop forwarding purposes
• Supports DHCPv6 prefix delegation (delegated prefixes move along with the MN)
• tunnels from MGs to MRs are bi-directional (but, tunneling from MR to MG can be omitted if “default” is propagated through MANET)