OSPF Route Summarization

What route summarization is for

Multiple areas and the ABR showed that splitting a network into areas still leaves an ABR originating one type 3 LSA per prefix. A hundred /32 loopbacks inside an area means a hundred type 3 LSAs flooding the next area — and a hundred routes in the routing table of every router that receives them. Summarization folds them into one.

The benefit is often described as a smaller LSDB, but what actually bites is the number of routes in the routing table (RIB) and the forwarding table (FIB). LSAs stay within OSPF process memory; route counts translate directly into forwarding hardware capacity and into how much of the table has to be rewritten. This article measures that.

OSPF can summarize in only two places: the ABR and the ASBR. Not inside an area, and not on a router in the middle of the backbone. Everything inside an area depends on every router holding the same LSDB, so dropping information halfway would make the calculations disagree.

Kind	Where it runs	IOS XR command	What it covers
Inter-area summarization	ABR	`range <prefix>` under `area <ID>`	Intra-area routes to type 3
External route summarization	ASBR	`summary-prefix <prefix>` under `router ospf`	Redistributed routes to type 5 / type 7

This article uses a seven-router, three-area lab to fold five loopbacks of area 2 into one route at the ABRs and four redistributed external routes into one at the ASBR, and checks on real devices how the metric of a summary LSA is chosen, what happens when the component routes go away, and what the side effects are.

Lab topology

The three-area lab of the stub area article is reused with the material to summarize added.

Item	Content
Intra-area routes to summarize	`10.2.0.1/32` and `10.2.0.2/32` on R11, `10.2.0.3/32` and `10.2.0.4/32` on R12, `10.2.0.5/32` on R13 (all in area 2, `passive enable`)
Summary at the ABRs	`10.2.0.0/24` on R3 and R4
External routes to summarize	`192.168.16.0/24`, `17.0/24`, `18.0/24` and `19.0/24`, redistributed by R1 as Null0 statics
Summary at the ASBR	`192.168.16.0/22` on R1
Left unsummarized	`192.168.1.0/24`, so the two can be compared
A real host inside the range	`192.168.16.1/32` on R1’s Loopback4, used as an address that exists inside the summary
Observation points	R2 (outside the area, receiving the summary) and R13 (inside the area, unaffected)

R11 is configured like this; the loopbacks to be summarized are in OSPF area 2 as well.

R11 show running-config (the router ospf 1 part, at boot)
router ospf 1
 router-id 11.11.11.11
 area 2
  interface Loopback0
   passive enable
  !
  interface Loopback2
   passive enable
  !
  interface Loopback3
   passive enable
  !
  interface GigabitEthernet0/0/0/0
   network point-to-point
  !
  interface GigabitEthernet0/0/0/1
   network point-to-point
  !
 !
!

Before summarization (STEP 0)

From R2 in area 0, the five /32 routes of area 2 are separate inter-area routes and R1’s four external routes are separate too. Note that the five costs vary: 3, 3, 5, 5 and 4 — that is the raw material for the summary metric.

STEP 0 R2 show route ospf (before summarization)
RP/0/RP0/CPU0:R2#show route ospf
Mon Sep  7 22:00:06.098 UTC

O    1.1.1.1/32 [110/2] via 10.0.12.1, 00:03:33, GigabitEthernet0/0/0/0
O    3.3.3.3/32 [110/2] via 10.0.23.3, 00:04:20, GigabitEthernet0/0/0/1
O    4.4.4.4/32 [110/2] via 10.0.24.4, 00:04:11, GigabitEthernet0/0/0/2
O IA 10.1.11.0/24 [110/3] via 10.0.23.3, 00:04:05, GigabitEthernet0/0/0/1
O IA 10.1.12.0/24 [110/4] via 10.0.23.3, 00:03:39, GigabitEthernet0/0/0/1
O IA 10.2.0.1/32 [110/3] via 10.0.23.3, 00:04:05, GigabitEthernet0/0/0/1
O IA 10.2.0.2/32 [110/3] via 10.0.23.3, 00:04:05, GigabitEthernet0/0/0/1
O IA 10.2.0.3/32 [110/5] via 10.0.23.3, 00:03:38, GigabitEthernet0/0/0/1
O IA 10.2.0.4/32 [110/5] via 10.0.23.3, 00:03:38, GigabitEthernet0/0/0/1
O IA 10.2.0.5/32 [110/4] via 10.0.23.3, 00:03:39, GigabitEthernet0/0/0/1
O IA 11.11.11.11/32 [110/3] via 10.0.23.3, 00:04:05, GigabitEthernet0/0/0/1
O IA 12.12.12.12/32 [110/5] via 10.0.23.3, 00:03:38, GigabitEthernet0/0/0/1
O IA 13.13.13.13/32 [110/4] via 10.0.23.3, 00:03:39, GigabitEthernet0/0/0/1
O IA 172.16.3.0/24 [110/2] via 10.0.23.3, 00:04:20, GigabitEthernet0/0/0/1
O IA 172.16.4.0/24 [110/11] via 10.0.24.4, 00:04:11, GigabitEthernet0/0/0/2
O E2 192.168.1.0/24 [110/20] via 10.0.12.1, 00:03:33, GigabitEthernet0/0/0/0
O E2 192.168.16.0/24 [110/20] via 10.0.12.1, 00:03:33, GigabitEthernet0/0/0/0
O E2 192.168.17.0/24 [110/20] via 10.0.12.1, 00:03:33, GigabitEthernet0/0/0/0
O E2 192.168.18.0/24 [110/20] via 10.0.12.1, 00:03:33, GigabitEthernet0/0/0/0
O E2 192.168.19.0/24 [110/20] via 10.0.12.1, 00:03:33, GigabitEthernet0/0/0/0

Reachability first: 10.2.0.1 (a loopback in area 2) and 192.168.16.1 (R1’s Loopback4) answer, while 192.168.19.1 (inside 192.168.19.0/24 but with nothing behind it) does not.

STEP 0 R2 ping (before summarization)
RP/0/RP0/CPU0:R2#ping 10.2.0.1 source 2.2.2.2
Mon Sep  7 21:59:23.400 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.2.0.1 timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 7/7/9 ms
RP/0/RP0/CPU0:R2#ping 192.168.16.1 source 2.2.2.2
Mon Sep  7 21:59:24.313 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 192.168.16.1 timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 4/4/7 ms
RP/0/RP0/CPU0:R2#ping 192.168.19.1 source 2.2.2.2
Mon Sep  7 21:59:25.296 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 192.168.19.1 timeout is 2 seconds:
.....
Success rate is 0 percent (0/5)

Summarizing at the ABR (STEP 1)

Both ABRs attached to area 2 (R3 and R4) get the 10.2.0.0/24 range.

The configuration applied in STEP 1
(on both R3 and R4)
router ospf 1
 area 2
  range 10.2.0.0/24

From R2 the five /32 routes are gone, replaced by a single 10.2.0.0/24. In the LSDB the summary LSA R3 originated has metric 4 and R4’s has metric 13.

STEP 1 R2 show route ospf and show ospf database summary 10.2.0.0 (first part)
RP/0/RP0/CPU0:R2#show route ospf
Mon Sep  7 22:03:40.100 UTC

O    1.1.1.1/32 [110/2] via 10.0.12.1, 00:07:07, GigabitEthernet0/0/0/0
O    3.3.3.3/32 [110/2] via 10.0.23.3, 00:07:54, GigabitEthernet0/0/0/1
O    4.4.4.4/32 [110/2] via 10.0.24.4, 00:07:45, GigabitEthernet0/0/0/2
O IA 10.1.11.0/24 [110/3] via 10.0.23.3, 00:07:39, GigabitEthernet0/0/0/1
O IA 10.1.12.0/24 [110/4] via 10.0.23.3, 00:07:13, GigabitEthernet0/0/0/1
O IA 10.2.0.0/24 [110/5] via 10.0.23.3, 00:00:31, GigabitEthernet0/0/0/1
O IA 11.11.11.11/32 [110/3] via 10.0.23.3, 00:07:39, GigabitEthernet0/0/0/1
O IA 12.12.12.12/32 [110/5] via 10.0.23.3, 00:07:12, GigabitEthernet0/0/0/1
O IA 13.13.13.13/32 [110/4] via 10.0.23.3, 00:07:13, GigabitEthernet0/0/0/1
O IA 172.16.3.0/24 [110/2] via 10.0.23.3, 00:07:54, GigabitEthernet0/0/0/1
O IA 172.16.4.0/24 [110/11] via 10.0.24.4, 00:07:45, GigabitEthernet0/0/0/2
O E2 192.168.1.0/24 [110/20] via 10.0.12.1, 00:07:07, GigabitEthernet0/0/0/0
O E2 192.168.16.0/24 [110/20] via 10.0.12.1, 00:07:07, GigabitEthernet0/0/0/0
O E2 192.168.17.0/24 [110/20] via 10.0.12.1, 00:07:07, GigabitEthernet0/0/0/0
O E2 192.168.18.0/24 [110/20] via 10.0.12.1, 00:07:07, GigabitEthernet0/0/0/0
O E2 192.168.19.0/24 [110/20] via 10.0.12.1, 00:07:07, GigabitEthernet0/0/0/0
RP/0/RP0/CPU0:R2#show ospf database summary 10.2.0.0
Mon Sep  7 22:03:40.301 UTC


            OSPF Router with ID (2.2.2.2) (Process ID 1)

		Summary Net Link States (Area 0)

  Routing Bit Set on this LSA
  LS age: 33
  Options: (No TOS-capability, DC)
  LS Type: Summary Links (Network)
  Link State ID: 10.2.0.0 (Summary Network Number)
  Advertising Router: 3.3.3.3
  LS Seq Number: 80000001
  Checksum: 0xbc64
  Length: 28
  Network Mask: /24
	TOS: 0 	Metric: 4 

  LS age: 18

How that metric is chosen changed between the two versions of the specification. R3’s costs to the components are 2 for 10.2.0.1, 2 for 10.2.0.2, 3 for 10.2.0.5 and 4 for both 10.2.0.3 and 10.2.0.4, so the summary metric of 4 is the largest of them.

Specification	Rule
RFC 2328 (current)	Section 12.4.3: “cost equal to the largest cost of any of the component networks”
RFC 1583 (superseded)	Section 12.4.3: “cost equal to the smallest cost of any of the component networks”
Measured on XRd 26.1.1	4, the largest — it follows RFC 2328

A summarizing ABR also installs a discard route to Null0. It drops packets for addresses that fall inside the range but do not actually exist, which is what keeps them from looping. Its administrative distance is 254, and this behaviour is an implementation choice, not something the RFC defines.

STEP 1 R3 show route | include Null0
RP/0/RP0/CPU0:R3#show route | include Null0
Mon Sep  7 22:03:43.840 UTC
O IA 10.2.0.0/24 [254/0] via 0.0.0.0, 00:00:35, Null0

The LS Updates that crossed the backbone (between R2 and R3) show the whole operation. No.4 of the attached capture is the summary LSA (metric 4) appearing, and No.5 is the LS Update that withdraws all five individual LSAs at once — done with LS Age 3600 (MaxAge) together with Metric: 16777215 (LSInfinity).

No.4 LS Update (the summary LSA appears) tshark -V
Open Shortest Path First
    OSPF Header
        Version: 2
        Message Type: LS Update (4)
        Packet Length: 56
        Source OSPF Router: 3.3.3.3
        Area ID: 0.0.0.0 (Backbone)
        Checksum: 0x8a29 [correct]
        Instance ID: Base IPv4 Unicast Instance (0)
        Auth Type: Null (0)
        Auth Data (none): 0000000000000000
    LS Update Packet
        Number of LSAs: 1
        LSA-type 3 (Summary-LSA (IP network)), len 28
            .000 0000 0000 0001 = LS Age (seconds): 1
            0... .... .... .... = Do Not Age Flag: 0
            Options: 0x22, (DC) Demand Circuits, (E) External Routing
                0... .... = DN: Not set
                .0.. .... = (O) Opaque: Not set
                ..1. .... = (DC) Demand Circuits: Supported
                ...0 .... = (L) LLS Data block: Not Present
                .... 0... = (N) NSSA: Not supported
                .... .0.. = (MC) Multicast: Not capable
                .... ..1. = (E) External Routing: Capable
                .... ...0 = (MT) Multi-Topology Routing: No
            LS Type: Summary-LSA (IP network) (3)
            Link State ID: 10.2.0.0
            Advertising Router: 3.3.3.3
            Sequence Number: 0x80000001
            Checksum: 0xbc64
            Length: 28
            Netmask: 255.255.255.0
            TOS: 0
            Metric: 4

Download the pcap of the packet in the tshark output above (No.4, the summary LSU)

The withdrawal carries five LSAs in one packet, as the tshark listing shows.

The LS Updates in the capture (tshark filtered on ospf.msg == 4)
13.044742    10.0.23.3 → 224.0.0.5    OSPF 90 LS Update
13.149976    10.0.23.3 → 224.0.0.5    OSPF 202 LS Update
27.555663    10.0.23.2 → 224.0.0.5    OSPF 90 LS Update
27.668405    10.0.23.2 → 224.0.0.5    OSPF 202 LS Update

Download the pcap of No.5 above (the LSU that withdraws all five individual LSAs)

When one component goes away (STEP 2)

This is where the effect of summarization is clearest. R11’s Loopback2 (10.2.0.1/32) is shut down.

The configuration applied in STEP 2
(R11)
interface Loopback2
 shutdown

From R2, 10.2.0.0/24 does not change at all, neither its cost nor the LSA. Outside the area nothing appears to have happened. A ping to 10.2.0.1 no longer works, though, while 10.2.0.2 still does: stability comes at the price of hiding reachability changes too.

STEP 2 R2 show route and show ospf database summary (first part)
RP/0/RP0/CPU0:R2#show route 10.2.0.0/24
Mon Sep  7 22:07:35.738 UTC

Routing entry for 10.2.0.0/24
  Known via "ospf 1", distance 110, metric 5, type inter area
  Installed Sep  7 22:03:08.925 for 00:04:26
  Routing Descriptor Blocks
    10.0.23.3, from 3.3.3.3, via GigabitEthernet0/0/0/1
      Route metric is 5
  No advertising protos. 
RP/0/RP0/CPU0:R2#show ospf database summary 10.2.0.0
Mon Sep  7 22:07:35.900 UTC


            OSPF Router with ID (2.2.2.2) (Process ID 1)

		Summary Net Link States (Area 0)

  Routing Bit Set on this LSA
  LS age: 268
  Options: (No TOS-capability, DC)
  LS Type: Summary Links (Network)
  Link State ID: 10.2.0.0 (Summary Network Number)
  Advertising Router: 3.3.3.3

STEP 2 R2 ping (the rest of the same output)
  LS age: 267
  Options: (No TOS-capability, DC)
  LS Type: Summary Links (Network)
  Link State ID: 10.2.0.0 (Summary Network Number)
  Advertising Router: 2.2.2.2
  LS Seq Number: 80000001
  Checksum: 0xe43f
  Length: 28
  Network Mask: /24
	TOS: 0 	Metric: 5 

RP/0/RP0/CPU0:R2#ping 10.2.0.1 source 2.2.2.2
Mon Sep  7 22:07:36.641 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.2.0.1 timeout is 2 seconds:
.....
Success rate is 0 percent (0/5)
RP/0/RP0/CPU0:R2#ping 10.2.0.2 source 2.2.2.2
Mon Sep  7 22:07:47.613 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.2.0.2 timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 8/8/10 ms

Inside the area the reaction is immediate. On R13, in the same area 2, the individual route disappears at once. Summarization only ever applies to what leaves the area.

STEP 2 R13 show route 10.2.0.1/32 (gone inside the area)
RP/0/RP0/CPU0:R13#show route 10.2.0.1/32
Mon Sep  7 22:07:51.571 UTC

% Network not in table

When every component goes away (STEP 3)

Shutting down the remaining four makes the summary LSA itself disappear.

The configuration applied in STEP 3
(R11's Loopback3, R12's Loopback2 and Loopback3, R13's Loopback2)
interface LoopbackN
 shutdown

STEP 3 R2 show route 10.2.0.0/24 and show ospf database summary 10.2.0.0
RP/0/RP0/CPU0:R2#show route 10.2.0.0/24
Mon Sep  7 22:12:03.931 UTC

% Network not in table

RP/0/RP0/CPU0:R2#show ospf database summary 10.2.0.0
Mon Sep  7 22:12:04.217 UTC


            OSPF Router with ID (2.2.2.2) (Process ID 1)

The ABR’s discard route goes with it. Section 12.4.3 of RFC 2328 says that when an advertised destination becomes unreachable the router must flush the LSA by setting its age to MaxAge, so the summary LSA lives as long as at least one component does, and is withdrawn only when the last one is gone.

STEP 3 R3 show route | include Null0 (the discard route is gone too)
RP/0/RP0/CPU0:R3#show route | include Null0
Mon Sep  7 22:12:08.043 UTC

Hiding a range instead of summarizing it (STEP 4)

A range can be given not-advertise: no summary LSA and no individual LSAs either, which hides the range completely from outside the area. The components are restored first.

The configuration applied in STEP 4
(after bringing the loopbacks of R11, R12 and R13 back with no shutdown)

(on both R3 and R4)
router ospf 1
 area 2
  range 10.2.0.0/24 not-advertise

STEP 4 R2 show route (neither the summary nor the components)
RP/0/RP0/CPU0:R2#show route 10.2.0.0/24
Mon Sep  7 22:16:44.339 UTC

% Network not in table

RP/0/RP0/CPU0:R2#show route 10.2.0.1/32
Mon Sep  7 22:16:44.515 UTC

% Network not in table

RP/0/RP0/CPU0:R2#ping 10.2.0.1 source 2.2.2.2
Mon Sep  7 22:16:44.696 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.2.0.1 timeout is 2 seconds:
.....
Success rate is 0 percent (0/5)

Inside the area everything is still reachable. Section 12.4.3 of RFC 2328 states that with DoNotAdvertise the type 3 LSA is suppressed and the component networks stay hidden from other areas.

STEP 4 R13 show route 10.2.0.1/32 (alive inside the area)
RP/0/RP0/CPU0:R13#show route 10.2.0.1/32
Mon Sep  7 22:16:58.253 UTC

Routing entry for 10.2.0.1/32
  Known via "ospf 1", distance 110, metric 2, type intra area
  Installed Sep  7 22:15:33.422 for 00:01:24
  Routing Descriptor Blocks
    10.1.11.11, from 11.11.11.11, via GigabitEthernet0/0/0/0
      Route metric is 2
  No advertising protos. 

Summarizing external routes at the ASBR (STEP 5)

After putting the range back to normal, it is the ASBR’s turn: R1 folds the four /24 routes it redistributes into 192.168.16.0/22.

The configuration applied in STEP 5
(R3 and R4 put the range back to normal)
router ospf 1
 area 2
  range 10.2.0.0/24

(R1)
router ospf 1
 summary-prefix 192.168.16.0/22

R2’s four external routes become one. 192.168.1.0/24, which is not summarized, stays as it is.

STEP 5 R2 show route ospf (the external routes are down to one)
RP/0/RP0/CPU0:R2#show route ospf
Mon Sep  7 22:20:53.996 UTC

O    1.1.1.1/32 [110/2] via 10.0.12.1, 00:24:21, GigabitEthernet0/0/0/0
O    3.3.3.3/32 [110/2] via 10.0.23.3, 00:25:08, GigabitEthernet0/0/0/1
O    4.4.4.4/32 [110/2] via 10.0.24.4, 00:24:59, GigabitEthernet0/0/0/2
O IA 10.1.11.0/24 [110/3] via 10.0.23.3, 00:24:53, GigabitEthernet0/0/0/1
O IA 10.1.12.0/24 [110/4] via 10.0.23.3, 00:24:27, GigabitEthernet0/0/0/1
O IA 10.2.0.0/24 [110/5] via 10.0.23.3, 00:00:42, GigabitEthernet0/0/0/1
O IA 11.11.11.11/32 [110/3] via 10.0.23.3, 00:24:53, GigabitEthernet0/0/0/1
O IA 12.12.12.12/32 [110/5] via 10.0.23.3, 00:24:26, GigabitEthernet0/0/0/1
O IA 13.13.13.13/32 [110/4] via 10.0.23.3, 00:24:27, GigabitEthernet0/0/0/1
O IA 172.16.3.0/24 [110/2] via 10.0.23.3, 00:25:08, GigabitEthernet0/0/0/1
O IA 172.16.4.0/24 [110/11] via 10.0.24.4, 00:24:59, GigabitEthernet0/0/0/2
O E2 192.168.1.0/24 [110/20] via 10.0.12.1, 00:24:21, GigabitEthernet0/0/0/0
O E2 192.168.16.0/22 [110/20] via 10.0.12.1, 00:00:13, GigabitEthernet0/0/0/0
RP/0/RP0/CPU0:R2#show ospf database external 192.168.16.0
Mon Sep  7 22:20:54.198 UTC

The ASBR gets a discard route as well. The S entries are the statics being redistributed; O E2 ... [254/0] ... Null0 is the discard route created by the summary.

STEP 5 R1 show route | include Null0
RP/0/RP0/CPU0:R1#show route | include Null0
Mon Sep  7 22:20:58.112 UTC
S    192.168.1.0/24 is directly connected, 00:24:46, Null0
O E2 192.168.16.0/22 [254/0] via 0.0.0.0, 00:00:17, Null0
S    192.168.16.0/24 is directly connected, 00:24:46, Null0
S    192.168.17.0/24 is directly connected, 00:24:46, Null0
S    192.168.18.0/24 is directly connected, 00:24:46, Null0
S    192.168.19.0/24 is directly connected, 00:24:46, Null0

The side effect: a black hole

Summarizing also draws packets for addresses that fall inside the range but do not exist all the way to the summarizing router, where Null0 drops them. Two pings from R13 make the difference obvious.

192.168.16.1 — exists as R1’s Loopback4, so it answers
192.168.19.1 — inside 192.168.19.0/24 with nothing behind it, so it does not (dropped by R1’s Null0)

STEP 5 R13 ping and traceroute (a non-existent address inside the range)
RP/0/RP0/CPU0:R13#ping 192.168.16.1 source 13.13.13.13
Mon Sep  7 22:25:40.945 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 192.168.16.1 timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 13/14/19 ms
RP/0/RP0/CPU0:R13#ping 192.168.19.1 source 13.13.13.13
Mon Sep  7 22:25:41.993 UTC
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 192.168.19.1 timeout is 2 seconds:
.....
Success rate is 0 percent (0/5)
RP/0/RP0/CPU0:R13#traceroute 192.168.19.1 source 13.13.13.13 maxttl 5
Mon Sep  7 22:25:53.044 UTC

Type escape sequence to abort.
Tracing the route to 192.168.19.1

 1  10.1.11.11 5 msec  62 msec  6 msec 
 2  172.16.3.3 8 msec  7 msec  7 msec 
 3  10.0.23.2 10 msec  10 msec  10 msec 
 4   *  *  * 
 5   *  *  * 

The traceroute reaches 10.0.23.2 (R2) and then goes quiet, because the packets get to R1 and die on Null0 (it is cut off with maxttl 5). The same address behaved the same way before summarization, of course; what changes is that after summarizing, every unused address in the whole range behaves like this.

Restoring (STEP 6)

Removing the range on the ABRs and the summary-prefix on the ASBR brings the individual routes back.

The configuration applied in STEP 6 (final state)
(R3 and R4)
router ospf 1
 area 2
  no range 10.2.0.0/24

(R1)
router ospf 1
 no summary-prefix 192.168.16.0/22

STEP 6 R2 show route ospf (after the restore)
RP/0/RP0/CPU0:R2#show route ospf
Mon Sep  7 22:27:30.191 UTC

O    1.1.1.1/32 [110/2] via 10.0.12.1, 00:30:57, GigabitEthernet0/0/0/0
O    3.3.3.3/32 [110/2] via 10.0.23.3, 00:31:44, GigabitEthernet0/0/0/1
O    4.4.4.4/32 [110/2] via 10.0.24.4, 00:31:35, GigabitEthernet0/0/0/2
O IA 10.1.11.0/24 [110/3] via 10.0.23.3, 00:31:29, GigabitEthernet0/0/0/1
O IA 10.1.12.0/24 [110/4] via 10.0.23.3, 00:31:03, GigabitEthernet0/0/0/1
O IA 10.2.0.1/32 [110/3] via 10.0.23.3, 00:00:52, GigabitEthernet0/0/0/1
O IA 10.2.0.2/32 [110/3] via 10.0.23.3, 00:00:52, GigabitEthernet0/0/0/1
O IA 10.2.0.3/32 [110/5] via 10.0.23.3, 00:00:52, GigabitEthernet0/0/0/1
O IA 10.2.0.4/32 [110/5] via 10.0.23.3, 00:00:52, GigabitEthernet0/0/0/1
O IA 10.2.0.5/32 [110/4] via 10.0.23.3, 00:00:52, GigabitEthernet0/0/0/1
O IA 11.11.11.11/32 [110/3] via 10.0.23.3, 00:31:29, GigabitEthernet0/0/0/1
O IA 12.12.12.12/32 [110/5] via 10.0.23.3, 00:31:02, GigabitEthernet0/0/0/1
O IA 13.13.13.13/32 [110/4] via 10.0.23.3, 00:31:03, GigabitEthernet0/0/0/1
O IA 172.16.3.0/24 [110/2] via 10.0.23.3, 00:31:44, GigabitEthernet0/0/0/1
O IA 172.16.4.0/24 [110/11] via 10.0.24.4, 00:31:35, GigabitEthernet0/0/0/2
O E2 192.168.1.0/24 [110/20] via 10.0.12.1, 00:30:57, GigabitEthernet0/0/0/0
O E2 192.168.16.0/24 [110/20] via 10.0.12.1, 00:00:21, GigabitEthernet0/0/0/0
O E2 192.168.17.0/24 [110/20] via 10.0.12.1, 00:00:21, GigabitEthernet0/0/0/0
O E2 192.168.18.0/24 [110/20] via 10.0.12.1, 00:00:21, GigabitEthernet0/0/0/0

How much the routing table shrinks

Summarization is usually described in terms of the LSDB holding fewer LSAs, but what matters in operation is that every router ends up with fewer entries in its routing table (RIB) and forwarding table (FIB). The LSA count is about OSPF process memory and the SPF run; the route count is about the capacity of the forwarding hardware (FIB or TCAM) and about how much has to be rewritten when something changes.

Counting the total of show route on R2 (area 0):

State	Total in `show route`	Of which OSPF	`/32` of area 2	External `/24`
No summarization (STEP 0)	27	20	5	5
Summarized at the ABR (STEP 1)	23	16	1 (summary)	5
`not-advertise` (STEP 4)	22	15	0	5
ABR and ASBR (STEP 5)	20	13	1 (summary)	2
Restored (STEP 6)	27	20	5	5

Folding five routes and four routes into one each takes the routing table from 27 entries down to 20, a 26% reduction. Real networks often carry hundreds or thousands of prefixes per area, and every one of them consumes FIB space on every router. Summarization cuts exactly that.

And it is not only the summarizing router that benefits: every router that receives the summarized LSA saves the same entries.

Router	Position	No summarization (STEP 0)	ABR and ASBR (STEP 5)
R2	Area 0, next to the ABRs	27	20
R1	Area 1, the ASBR itself	27	24
R13	Inside area 2	26	23

R13 sits inside area 2, so it still holds the 10.2.0.x/32 routes individually, but summarizing the external routes still saves it three entries. Summarization shrinks the tables of every router downstream of it.

As a secondary benefit, a prefix flapping inside the area no longer changes the LSDB outside it (STEP 2), so SPF stops running outside the area.

Design notes

It depends on an addressing plan that can be summarized. Without a contiguous block per area, nothing folds into one. Since OSPF can only summarize at an ABR or an ASBR, retrofitting it later is painful
The metric is the largest of the components (RFC 2328). The cost of the most distant component becomes the representative value, so a summary can make the area look farther away than it is
Unused addresses inside the range become a black hole, dropped by the summarizing router’s Null0. That is deliberate loop prevention, but during troubleshooting it looks like “the route is there yet nothing gets through”, so keep the range in mind
not-advertise works as a filter when a range should not leave the area at all — at the cost of reachability to it
Stub areas and NSSAs solve a different problem. Stub areas and NSSAs keep LSAs out; summarization folds them together. The two combine well

The IOS XR commands

Command	Where it goes	Meaning
`range <prefix>`	Under `area <ID>` on an ABR	Folds that area’s intra-area routes into a single type 3
`range <prefix> not-advertise`	Under `area <ID>` on an ABR	Hides them instead of folding them (no type 3 at all)
`summary-prefix <prefix>`	Under `router ospf` on an ASBR	Folds redistributed external routes into a single type 5

Example (ABR R3 and ASBR R1)
(the ABR)
router ospf 1
 area 2
  range 10.2.0.0/24
 !
!

(the ASBR)
router ospf 1
 summary-prefix 192.168.16.0/22
!

A range goes on every ABR attached to that area. Miss one and individual type 3 LSAs keep flowing through it, defeating the purpose. On IOS XE the equivalents are area <ID> range <address> <mask> and summary-address <address> <mask>. Configuration examples are covered in a separate article.

Verification config and show output

For each STEP the following three files were collected from all seven routers, split per router. The verification config is the ..._run.txt file (the final state is the one from STEP 6).

File	Content
`..._show.txt`	`show version` / `show interface description` / `show route` / `show route ospf` / `show ospf` / `show ospf interface` / `show ospf interface brief` / `show ospf neighbor` / `show ospf neighbor detail` / `show ospf database` / `show ospf database router` / `show ospf database network` / `show ospf statistics interface` / `show ospf database summary` / `show ospf database external` / `show ospf border-routers` / `show ospf routes` / `show route 10.2.0.0/24` / `show route 10.2.0.1/32` / `show route 192.168.16.0/22` / `show route 192.168.16.0/24` / `show route 192.168.1.0/24` / `show ospf trace events`
`..._log.txt`	`show logging` narrowed to that STEP. A marker was written with `logmsg` at the start of each STEP and its timestamp passed to `show logging start`
`..._run.txt`	`show running-config` at that point in time (that is, the verification config for that STEP)

The final state has all summarization removed.

STEP 0: no summarization — the five /32 routes of area 2 and the four external routes all flow individually

Router	show output	syslog	running-config
R1	show	log	run
R2	show	log	run
R3	show	log	run
R4	show	log	run
R11	show	log	run
R12	show	log	run
R13	show	log	run

This STEP also has a ping from R2 and the Null0 check on R3.

STEP 1: area 2 range 10.2.0.0/24 on R3 and R4 — five routes become one, with the summary metric at the largest component cost (4)

Router	show output	syslog	running-config
R1	show	log	run
R2	show	log	run
R3	show	log	run
R4	show	log	run
R11	show	log	run
R12	show	log	run
R13	show	log	run

This STEP also has show route ospf and show ospf database summary on R2, the Null0 check on R3, a ping on R2.

STEP 2: R11’s Loopback2 is shut down — the summary LSA does not change, but that destination stops answering

Router	show output	syslog	running-config
R1	show	log	run
R2	show	log	run
R3	show	log	run
R4	show	log	run
R11	show	log	run
R12	show	log	run
R13	show	log	run

This STEP also has show route, database and ping on R2, the view from inside the area on R13.

STEP 3: all five components are shut down — the summary LSA and the Null0 discard route disappear

Router	show output	syslog	running-config
R1	show	log	run
R2	show	log	run
R3	show	log	run
R4	show	log	run
R11	show	log	run
R12	show	log	run
R13	show	log	run

This STEP also has show route and database on R2, the Null0 check on R3.

STEP 4: the components are restored and range ... not-advertise is applied — nothing is advertised, hiding the range

Router	show output	syslog	running-config
R1	show	log	run
R2	show	log	run
R3	show	log	run
R4	show	log	run
R11	show	log	run
R12	show	log	run
R13	show	log	run

This STEP also has show route on R2, the view from inside the area on R13.

STEP 5: the range is restored and R1 gets summary-prefix 192.168.16.0/22 — four external routes become one

Router	show output	syslog	running-config
R1	show	log	run
R2	show	log	run
R3	show	log	run
R4	show	log	run
R11	show	log	run
R12	show	log	run
R13	show	log	run

This STEP also has show route ospf and show ospf database external on R2, the Null0 check on R1, ping and traceroute on R13.

STEP 6: the range and the summary-prefix are removed (final state) — the individual routes are back

Router	show output	syslog	running-config
R1	show	log	run
R2	show	log	run
R3	show	log	run
R4	show	log	run
R11	show	log	run
R12	show	log	run
R13	show	log	run

This STEP also has show route ospf and ping on R2.

The capture was taken on the backbone side (between R2 and R3) across the summarization of STEP 1, filtered with ip proto 89.

Download the capture of the ABR summarization (STEP 1)

References

RFC	Title	Summary
RFC 2328	OSPF Version 2	Area address ranges (section 3.5); originating summary-LSAs for a range, the cost being the largest of the components, DoNotAdvertise and the MaxAge flush when a destination becomes unreachable (section 12.4.3).
RFC 1583	OSPF Version 2 (superseded)	The same section 12.4.3 defined the cost as the smallest of the components; RFC 2328 changed it to the largest.