Automatic Subnet Numbering

1
Automatic Subnet Numbering

• Hilarie Orman

2
Layer Addressing Characteristics

• Layer 2
• Unique, uncorrelated addresses
• Linear strings of nodes
• Join points
• Spanning tree
• Layer 3
• Address blocks grouped by contiguous bits
• Mesh of routers
• Configured topology over layer 2
• Universal address space

3
Routers and Subnetswithout IP addrs, no routing
X
X
X
X
4
IP Address
Subnet number
Fixed prefix ARIN or non-routable
Host number DHCP
5
SubnetNumbers
SUBNET 7
SUBNET 1
X
SUBNET8
SUBNET 5
SUBNET 2
SUBNET 9
X
X
SUBNET 3
SUBNET 6
X
SUBNET 4
6
Requirements

• Support initialize, join, leave, partition,
  rejoin
• Utilize IPv4 addresses efficiently
• Solve v6 and v4 assignments
• Cooperate with DHCP
• Minimal interference with ongoing end node
  connections
• Be secure

7
Central Authority
Request Subnet Number
X
Reply Subnet Number
Single point of failure --- ameliorate with
elections
8
Distributed Agreement Protocol
X
X
X
Router 1 needs 3 subnet numbers
X
Router 2 needs 2 subnet numbers
X
9
Distributed Protocol Problems

• Might be slow at startup
• Partitions due to failure require renumbering
• Partition joins as part of network growth would
  be complicated
• Authentication via secure multicast keying?

10
Network Partition / Nodes Join
SUBNET 7
SUBNET 1
SUBNET 10
X
SUBNET8
SUBNET 2
SUBNET 9
X
X
SUBNET 3
SUBNET 10
X
SUBNET 4
11
Node Mobility
SUBNET 1
Prefix subnet 1 hostnum
X
SUBNET 2
SUBNET 4
X
SUBNET 3
Prefix subnet 4 hostnum
12
Address Utilization

• Mobility makes subnet size difficult to predict
• Simplest assignment method divides address space
  into N subnets of equal size
• Must overprovision and waste
• Dynamic subnet size adjustment
• Difficult to synchronize

13
Automatic Numbering at Layer 3Main Points

• No end node mobility without mobile IP
• Partition/rejoin might require reassigning end
  node addresses
• Breaks connections
• Sizing subnets is awkward