Skip to content
Log in Sign up My account

Case 02: Multi-Area Routing and BGP Edge Design

Blueprint 1.2, 1.4, 1.5.a, 1.5.b, 1.5.c, 1.5.d, 1.5.e | Difficulty: Hard

Company background

Cobalt Steelworks runs 14 plants and two data centers across one country. Production scheduling depends on the network. The routing design has not changed since the company had four plants.

Current network

All 600 routers run OSPF in a single area 0. A flapping interface at one plant triggers SPF runs on every router. The two data centers each connect to a different ISP over eBGP (AS 64501 and AS 64502 are the upstream ISPs, the company uses AS 65010). Each data center border router has full-mesh iBGP with the other 11 BGP-speaking routers in the company.

graph TB
    ISPA(("ISP A<br/>AS 64501"))
    ISPB(("ISP B<br/>AS 64502"))
    DC1["DC1 border<br/>AS 65010"]
    DC2["DC2 border<br/>AS 65010"]
    CORE["OSPF area 0<br/>600 routers"]
    P["14 plants"]
    ISPA --- DC1
    ISPB --- DC2
    DC1 --- CORE
    DC2 --- CORE
    CORE --- P

Requirements

Business:

  • A plant instability must not affect other plants.
  • Internet traffic should leave through ISP A normally and fail over to ISP B.
  • Customers should reach company web servers through ISP A normally.

Technical:

  • Routing convergence inside a region should stay under a second.
  • The company must not carry traffic between its two ISPs.
  • iBGP configuration must scale to 30 BGP speakers without a full mesh.

Constraints

  • Both ISPs allow communities and accept more-specific prefixes down to /24.
  • The company has one /22 of public address space.
  • The OSPF design may be changed. A move to IS-IS is under consideration for the backbone only if it offers a clear benefit.

Design questions

1. Blueprint 1.4 What is the best change to the OSPF design?

  • A. Keep area 0 and add ip ospf cost changes on plant links.
  • B. Create one non-backbone area per region or plant group, keep area 0 for the DC and core routers, summarise at the ABRs, and make plant areas stubs or totally stubby where they have no external routes.
  • C. Split into many OSPF processes and redistribute between them at every router.
  • D. Replace OSPF with static routes at the plants.
Answer: B

Areas contain LSA flooding and SPF runs. A flap inside one area is hidden from the rest by the ABR, and summarisation means the summary does not change when an individual route flaps. Totally stubby or stub areas shrink the LSDB at the plants. A changes path choice but does not limit the failure domain. C creates many redistribution points and loop risk and does not provide a hierarchy. D removes dynamic failover.

2. Blueprint 1.2 The architects consider IS-IS for the backbone. Which statement about IS-IS design is correct?

  • A. The backbone is a contiguous set of L2-capable routers, L1 routers send unknown destinations to the nearest L1/L2 router, and area borders are on links instead of on routers.
  • B. IS-IS needs an area 0, and every area must touch it.
  • C. IS-IS runs over IP protocol 89.
  • D. IS-IS has no hierarchy and is flat.
Answer: A

IS-IS uses Level 1 inside an area and Level 2 between areas. A router can be L1, L2 or L1/L2, and an L1 router follows the attached bit toward an L1/L2 router for anything outside the area. The adjacency level, not the router, defines the border. B describes OSPF. C is wrong: IS-IS runs directly over Layer 2 (CLNS), not over IP. D is wrong: IS-IS has a two-level hierarchy. IS-IS is a good choice for a flat backbone with many routers or for an SD-Access underlay, but there is no clear benefit here that justifies a protocol migration for the plants, so the area redesign in question 1 is enough.

3. Blueprint 1.5.c How should the design make ISP A the preferred exit and also the preferred entry for customer traffic, with ISP B as backup in both directions?

  • A. Set a higher local preference on routes learned from ISP A for outbound, and advertise the /22 to ISP B with AS-path prepending (and with communities requesting lower provider local preference if the ISP supports it).
  • B. Set a higher MED toward ISP A.
  • C. Set a lower weight on routes from ISP A on both border routers.
  • D. Prepend the AS path on routes learned from ISP A.
Answer: A

Local preference is shared across the AS through iBGP and controls outbound path choice, so the highest local preference for ISP A routes makes both data centers send traffic out through A. Inbound path choice is controlled by what the company advertises: prepending the AS path toward ISP B makes the path through B look longer to the rest of the internet. B is wrong because MED is a hint to a neighbouring AS about how to reach you, and a higher MED is less preferred. C reverses the preference and weight is local to one router only. D prepends on received routes, which is meaningless for a path you advertise, and it would not control the company's traffic.

4. Blueprint 1.5.d The company expects 30 BGP speakers. Which design scales iBGP and keeps it resilient?

  • A. Keep the full mesh and add peer groups.
  • B. Two route reflectors in different data centers using the same cluster ID, with all other routers as clients.
  • C. One route reflector in DC1 only.
  • D. Use eBGP between all routers with private AS numbers.
Answer: B

Route reflectors remove the full mesh (n(n-1)/2 sessions) so each client peers with two reflectors. Two reflectors in separate sites avoid a single point of failure, and a shared cluster ID lets the reflectors discard each other's reflected routes. A only simplifies configuration and keeps the session count. C creates a single point of failure. D introduces a new AS per router and complicates policy, though confederations are a valid but heavier alternative.

5. Blueprint 1.5.b How do you stop the company from becoming a transit network between ISP A and ISP B?

  • A. Advertise only the company's own prefixes to each ISP, using an outbound prefix-list (and an AS-path filter that permits only locally originated routes).
  • B. Block all traffic with an interface ACL.
  • C. Set no synchronization.
  • D. Use maximum-prefix on the session.
Answer: A

By default a BGP speaker will pass on eBGP-learned routes to other eBGP peers. An outbound filter that permits only the company's own prefix and locally originated routes prevents advertising ISP A's routes to ISP B and vice versa. B drops data traffic but still advertises the routes. C is obsolete and unrelated. D protects the company from a flood of routes from the ISP, not from re-advertising routes.

6. Blueprint 1.5.e, 1.5.a A plant uses two links to ISP A and wants to share load across them. It also needs the same design for IPv6. Which two statements are correct? (Choose two.)

  • A. eBGP multipath (maximum-paths) installs more than one path when the attributes tie on weight, local preference, AS-path length, origin, MED and IGP metric.
  • B. BGP load-shares across any two paths automatically.
  • C. Multiprotocol BGP can carry both IPv4 and IPv6 prefixes in one session, with each address family activated for the neighbour.
  • D. IPv6 requires a separate BGP AS.
  • E. Load sharing is only possible with iBGP.
Answer: A, C

BGP selects a single best path unless multipath is configured, and the paths must be equal in the best-path attributes. Address families let a single session carry IPv4 and IPv6 or separate sessions carry one each. B is wrong because BGP picks one best path by default. D and E are wrong: IPv6 uses the same AS, and eBGP multipath exists. If the two links went to different ISPs, bestpath as-path multipath-relax would be needed.

Break area 0 into a core of data center and distribution routers with non-backbone areas per plant group. Summarise on the ABRs. Keep OSPF for now. Use two route reflectors with a shared cluster ID. Prefer ISP A with local preference, prepend toward ISP B, filter the outbound advertisements, and use eBGP multipath on dual links.

graph TB
    ISPA(("ISP A"))
    ISPB(("ISP B"))
    DC1["DC1 border<br/>RR1"]
    DC2["DC2 border<br/>RR2"]
    A0["Area 0<br/>DC and distribution"]
    A1["Area 1<br/>plants 1 to 5"]
    A2["Area 2<br/>plants 6 to 10"]
    A3["Area 3<br/>plants 11 to 14"]
    ISPA -- "local pref high" --- DC1
    ISPB -- "prepend, backup" --- DC2
    DC1 --- A0
    DC2 --- A0
    A0 -- "ABR summary" --- A1
    A0 -- "ABR summary" --- A2
    A0 -- "ABR summary" --- A3