MOOSE Multi level Origin Organised Scalable Ethernet draft malc - - PowerPoint PPT Presentation

moose
SMART_READER_LITE
LIVE PREVIEW

MOOSE Multi level Origin Organised Scalable Ethernet draft malc - - PowerPoint PPT Presentation

MOOSE Multi level Origin Organised Scalable Ethernet draft malc armd moose 00 Malcolm Scott University of Cambridge Computer Laboratory Malcolm Scott draft-malc-armd-moose-00 1 Aim: Hierarchical MAC address space


slide-1
SLIDE 1

Malcolm Scott draft-malc-armd-moose-00 1

MOOSE

Multi‐level Origin‐Organised Scalable Ethernet

draft‐malc‐armd‐moose‐00

Malcolm Scott

University of Cambridge Computer Laboratory

slide-2
SLIDE 2

Malcolm Scott draft-malc-armd-moose-00 2

Aim: Hierarchical MAC address space

  • Current Ethernet: manufacturer‐assigned MAC address

valid anywhere on any network

– But every switch must store the location of every host

  • Hierarchical MAC addresses: address depends on

location e.g. [switch ID].[port ID].[host ID]

– Route frames according to hierarchy – Small forwarding databases – Run a routing protocol between switches

  • One “subnet” per switch – e.g. “02:11:11:00:00:00/24”
  • Don’t advertise individual MAC addresses (cf. TRILL Rbridges)
  • LAAs? High administrative overhead. So, instead...:
slide-3
SLIDE 3

Malcolm Scott draft-malc-armd-moose-00 3

MOOSE

  • “NAT for Ethernet”

– Dynamically allocate hosts hierarchical addresses – Perform source MAC address rewriting on ingress – No encapsulation: no costly rewriting of dest address – Looks like Ethernet from outside: transparent to hosts – We have an OpenFlow implementation

slide-4
SLIDE 4

Malcolm Scott draft-malc-armd-moose-00 4

Beyond simple protocols

  • Some protocols must be rewritten by switches

– Anything which puts MAC address in payload – ARP, DHCP: trivial for switches to deal with

  • Broadcast: unfortunate legacy

– Propagate broadcast traffic using reverse path forwarding (PIM): no explicit spanning tree protocol

  • Multicast and anycast for free

– (if we use a suitable routing protocol) – May be able to convert broadcast into multicast by inferring groups (e.g. DHCP servers) – see SEATTLE

slide-5
SLIDE 5

Malcolm Scott draft-malc-armd-moose-00 5

This is ongoing research; comments very welcome

This was a very brief overview: much more detail in draft‐malc‐armd‐moose‐00

Malcolm Scott

Malcolm.Scott@cl.cam.ac.uk http://www.cl.cam.ac.uk/~mas90/MOOSE/

slide-6
SLIDE 6

Malcolm Scott draft-malc-armd-moose-00 6

  • Spare slides follow
slide-7
SLIDE 7

Malcolm Scott draft-malc-armd-moose-00 7

Mobility

  • If a host moves, it is

allocated a new MAC address by its new switch

  • Other hosts may have the
  • ld address in ARP caches

1. Forward frames, IP Mobility style

(new switch discovers host’s old location by querying other switches for its real MAC address)

2. Gratuitous ARP, Xen VM migration style

Host B Host A

h

  • s

t r e l

  • c

a t e d t

  • n

e w s w i t c h

  • data forwarded

by ca re-of switch gratuitous AR P se nt by new home swi tch

slide-8
SLIDE 8

Malcolm Scott draft-malc-armd-moose-00 8

Allocation of host identifiers

  • Only the switch which allocates a host ID ever

uses it for forwarding

– More distant switches just use the switch ID

  • Therefore the detail of how host IDs are allocated

can vary between switches

– Sequential assignment – Port number and sequential portion (reduces address exhaustion attacks) – Hash of manufacturer‐assigned MAC address (deterministic: recoverable after crash)

slide-9
SLIDE 9

Malcolm Scott draft-malc-armd-moose-00 9

Security and isolation benefits

  • The number of switch IDs is more predictable

by the network admin than the number of MAC addresses

– Address flooding attacks are ineffective

  • Host‐specified MAC address

is not used for switching

– Spoofing is ineffective