If you're driving from San Francisco to Los Angeles, there are multiple routes: the scenic Pacific Coast Highway, the fast Interstate 5, or back roads through farmland. Each has different characteristics—some faster, some shorter in distance, some more reliable, some more beautiful. How do you choose?

Routers face the same fundamental challenge. Between any two points in a network, there are often multiple paths. Some have more bandwidth, some have lower latency, some traverse fewer hops, some are more reliable. The routing metric is how routers quantify these path characteristics and make optimal selections.

Metrics are the numerical language routers use to compare paths. Understanding metrics is understanding how and why your traffic takes the route it does—and how you can influence those decisions through network design and configuration.

What is a Routing Metric?

A routing metric is a numerical value that represents the quality, cost, or desirability of a particular network path. Routers use metrics to compare routes and select the best path to a destination.

A routing metric quantifies path characteristics, enabling routers to make consistent, deterministic decisions when multiple paths exist to the same destination.

Core Principles

1. Lower is typically better: Most protocols prefer routes with lower metrics (cost)
2. Within protocol comparison: Metrics are compared only between routes from the same protocol
3. Deterministic selection: Same inputs always produce same route selection
4. Additive nature: Most metrics accumulate along the path (each hop adds its cost)

The Role of Metrics in Path Selection

When multiple routes exist to the same destination from the same routing protocol:

Route selection process:

1. Both routes to 10.0.0.0/8 from OSPF:
   Path A: Metric 50
   Path B: Metric 75
   
2. Compare metrics (both from OSPF, same AD):
   50 < 75
   
3. Result: Path A selected (lower metric)

4. Install Path A in FIB

Note: If Path A metric = Path B metric = 50:
→ Equal-Cost Multipath (ECMP)
→ Both paths installed and used

Administrative Distance vs. Metric

These often get confused:

Example: A route to 10.0.0.0/8 exists from both OSPF (AD=110) and BGP (AD=20). First, AD wins—BGP route is used. Only if there were multiple BGP routes would the BGP metric come into play.

Different routing protocols use different metrics, each capturing different aspects of path quality.

Hop Count

The simplest metric: count the number of routers (hops) between source and destination.

Hop Count Metric:

Source ──[R1]──[R2]──[R3]── Destination
          1     2     3
          
Metric = 3 (three routers traversed)

Advantages:
- Extremely simple to calculate
- Low overhead (just count hops)
- Easy to understand

Disadvantages:
- Ignores link speed (a 10Mbps hop = a 10Gbps hop)
- Ignores congestion
- Ignores reliability
- May choose longer path in miles if fewer routers

Used by: RIP (maximum 15 hops, 16 = unreachable)

Bandwidth-Based Metrics

Used by OSPF and EIGRP to prefer higher-capacity links:

OSPF Cost Formula:
                    Reference Bandwidth
OSPF Cost = ──────────────────────────────
                    Interface Bandwidth

Default reference bandwidth: 100 Mbps (10^8 bps)

Examples (default reference):
┌─────────────────────────────────────────────┐
│ Interface Speed │ Calculation    │ Cost    │
├─────────────────────────────────────────────┤
│ 10 Mbps         │ 100/10         │ 10      │
│ 100 Mbps        │ 100/100        │ 1       │
│ 1 Gbps          │ 100/1000       │ 1 *     │
│ 10 Gbps         │ 100/10000      │ 1 *     │
└─────────────────────────────────────────────┘
* Problem: All Gigabit+ links have same cost!

Solution: Increase reference bandwidth
router ospf 1
  auto-cost reference-bandwidth 10000
  
With 10 Gbps reference:
│ 1 Gbps          │ 10000/1000     │ 10      │
│ 10 Gbps         │ 10000/10000    │ 1       │

Delay-Based Metrics

EIGRP includes delay in its calculation:

EIGRP Delay Component:
- Configured in tens of microseconds
- Cumulative: adds up along the path
- Reflects propagation and serialization delay

Typical delay values:
- Ethernet: 100 (1,000 microseconds)
- FastEthernet: 100 (1,000 microseconds)
- GigabitEthernet: 10 (100 microseconds)
- Serial T1: 20,000 (200,000 microseconds)
- Satellite: Very high (>500,000 microseconds)

Each major routing protocol has its own approach to metrics. Understanding these is essential for network design and troubleshooting.

RIP Metric (Hop Count)

RIP Metric Rules:
- Directly connected: Metric 1
- Each additional hop: +1
- Maximum: 15 (16 = infinity/unreachable)
- All links weighted equally regardless of speed

Example:
Network A ──[56kbps]── R1 ──[1Gbps]── R2 ──[1Gbps]── Dest
Network B ──[1Gbps]── R3 ──[1Gbps]── R4 ──[1Gbps]── R5 ──[1Gbps]── Dest

RIP chooses: Path A (3 hops) over Path B (4 hops)
Despite Path A having a 56Kbps bottleneck!

This is why RIP is rarely used today.

OSPF Metric (Cost)

OSPF Cost Calculation:
- Cost per interface based on bandwidth
- Path cost = sum of all interface costs along path
- Preference: Lower total cost wins

Path Example:
                    Cost: 10          Cost: 1           Cost: 1
Source ────[10Mbps]──── R1 ────[100Mbps]──── R2 ────[100Mbps]──── Dest
       
                    Cost: 1           Cost: 1
Source ────[100Mbps]──── R3 ────[100Mbps]──── Dest

Path 1 total cost: 10 + 1 + 1 = 12
Path 2 total cost: 1 + 1 = 2

OSPF chooses Path 2 (cost 2 < cost 12)

Manual Cost Override:
interface GigabitEthernet0/0
  ip ospf cost 50

EIGRP Metric (Composite)

EIGRP uses a composite metric calculated from multiple path characteristics:

EIGRP Classic Metric Formula:

Metric = [K1×Bandwidth + K2×Bandwidth/(256-Load) + K3×Delay] × [K5/(Reliability+K4)]

Default K values: K1=1, K2=0, K3=1, K4=0, K5=0

Simplified (with defaults):
Metric = [Bandwidth + Delay] × 256

Where:
- Bandwidth = 10^7 / minimum_bandwidth_in_kbps
- Delay = sum of delays in tens of microseconds

Example Path:
Link 1: 100 Mbps, delay 100 (1ms)
Link 2: 10 Mbps, delay 1000 (10ms)

Bandwidth component: 10^7 / 10,000 = 1000
Delay component: 100 + 1000 = 1100

Metric = (1000 + 1100) × 256 = 537,600

BGP Metric (Multi-Factor Path Selection)

BGP doesn't use a simple numeric metric. Instead, it follows a multi-step path selection process:

BGP Best Path Selection (Simplified):

1. Highest Weight (Cisco local preference override)
2. Highest LOCAL_PREF (prefer certain exit points)
3. Locally originated routes preferred
4. Shortest AS_PATH (fewer autonomous systems)
5. Lowest ORIGIN type (IGP < EGP < Incomplete)
6. Lowest MED (Multi-Exit Discriminator)
7. eBGP over iBGP
8. Lowest IGP metric to next-hop
9. Oldest route (for stability)
10. Lowest router-id (tiebreaker)

Note: 'Metric' in BGP output typically refers to MED,
but actual path selection involves all these factors.

BGP's approach reflects its use for policy-based routing between organizations, not just finding the 'shortest' path.

Composite metrics combine multiple path characteristics into a single value. This approach provides more nuanced path selection but adds complexity.

EIGRP Wide Metrics

Classic EIGRP metrics had scaling problems with high-speed links. EIGRP Wide Metrics address this:

Problem with Classic EIGRP:
- Metric uses 32-bit arithmetic
- Very high-speed links produce same metric
- Limited differentiation above 10Gbps

EIGRP Wide Metrics (EIGRP Named Mode):
- Uses 64-bit arithmetic
- Introduces new components
- Better scaling for modern networks

Wide Metric Calculation:

                   Throughput        Latency
Metric = ─────────────────────── × ────────────
              Minimum Bandwidth      K values
         
         × Reliability × Load × Hop adjustment × Energy

New Named Mode Configuration:
router eigrp MYNET
  address-family ipv4 unicast autonomous-system 100
    metric rib-scale 128  ! Converts to 32-bit for RIB

IS-IS Metrics

IS-IS uses configurable costs similar to OSPF:

IS-IS Metric Types:

1. Narrow Metrics (original):
   - 6-bit per link (0-63)
   - 10-bit path cost (0-1023)
   - Too small for modern networks

2. Wide Metrics (modern):
   - 24-bit per link (0-16777215)
   - 32-bit path cost (0-4294967295)
   - Required for any serious deployment

Configuration:
interface GigabitEthernet0/0
  isis metric 1000

router isis
  metric-style wide

Network engineers often need to influence path selection for policy or performance reasons. Metric manipulation is a primary tool for this.

OSPF Cost Manipulation

Scenario: Force traffic away from a specific link

Before:
R1 ──[Cost 10]── R2 ──[Cost 10]── Dest   = Total 20
R1 ──[Cost 10]── R3 ──[Cost 10]── Dest   = Total 20 (ECMP)

Goal: Prefer path through R2, use R3 as backup

Solution 1: Increase cost on R3 path
On R1 or R3:
  interface GigabitEthernet0/1  ! Link to R3
    ip ospf cost 100

After:
R1 ──[Cost 10]── R2 ──[Cost 10]── Dest   = Total 20 ✓
R1 ──[Cost 100]── R3 ──[Cost 10]── Dest  = Total 110 (backup)

Solution 2: Decrease cost on preferred path
On R1 or R2:
  interface GigabitEthernet0/0  ! Link to R2
    ip ospf cost 1

After:
R1 ──[Cost 1]── R2 ──[Cost 10]── Dest    = Total 11 ✓
R1 ──[Cost 10]── R3 ──[Cost 10]── Dest   = Total 20 (backup)

EIGRP Offset List

Offset lists add to or subtract from EIGRP metrics:

! Add 1000 to metric for routes learned from 10.1.1.1
access-list 1 permit 10.0.0.0 0.255.255.255

router eigrp 100
  offset-list 1 in 1000 GigabitEthernet0/0

Effect:
- Routes matching ACL 1 (10.0.0.0/8)
- Received on Gi0/0
- Have 1000 added to their metric
- Makes them less preferred than other paths

BGP LOCAL_PREF for Outbound Traffic

! Prefer ISP1 over ISP2 for all outbound traffic

route-map PREFER_ISP1 permit 10
  set local-preference 200
  
route-map PREFER_ISP2 permit 10
  set local-preference 100  ! Lower = less preferred

router bgp 65000
  neighbor 10.1.1.1 route-map PREFER_ISP1 in  ! ISP1
  neighbor 10.2.2.2 route-map PREFER_ISP2 in  ! ISP2

Result:
- Routes from ISP1: LOCAL_PREF 200 (preferred)
- Routes from ISP2: LOCAL_PREF 100 (backup)
- All outbound traffic exits via ISP1

One of the most challenging aspects of multi-protocol networks is that metrics from different protocols are not directly comparable.

The Incomparability Problem

Scenario:
Route to 10.0.0.0/8 from OSPF: Metric 100
Route to 10.0.0.0/8 from EIGRP: Metric 2816000
Route to 10.0.0.0/8 from BGP: Metric 0 (MED)

Question: Which has the 'best' metric?
Answer: IMPOSSIBLE to compare directly!

- OSPF 100 = Sum of bandwidth-inverse costs
- EIGRP 2816000 = Composite of many factors
- BGP 0 = MED hint (not even primary selection criteria)

The numbers mean completely different things.

Route Redistribution Metric Assignment

When redistributing between protocols, you must assign a metric in the destination protocol's terms:

! Redistributing OSPF into EIGRP
router eigrp 100
  redistribute ospf 1 metric 10000 1000 255 1 1500
                      │      │     │   │  │
                      │      │     │   │  └─ MTU (bytes)
                      │      │     │   └─ Load (1-255)
                      │      │     └─ Reliability (0-255)
                      │      └─ Delay (in tens of μs)
                      └─ Bandwidth (Kbps)

! Redistributing EIGRP into OSPF
router ospf 1
  redistribute eigrp 100 subnets metric 100 metric-type 1
                                 │
                                 └─ OSPF cost (single number)

The original EIGRP composite metric is lost;
a single OSPF cost must be chosen by the engineer.

Default Redistribution Metrics

Some protocols have default redistribution metrics:

Important: 'Infinite' in EIGRP redistribution means routes won't be redistributed unless you explicitly set a metric!

In production networks, metric decisions must consider factors beyond protocol specifications.

Metric Stability vs. Accuracy

Should metrics reflect real-time conditions?

Load-Sensitive Routing (Using EIGRP K2):

Pros:
- Routes adapt to congestion
- Optimal path changes as load changes
- Maximum bandwidth utilization

Cons:
- Route instability (constant recalculation)
- Oscillation (traffic moves, load moves, traffic moves back)
- CPU overhead for metric calculation
- Convergence delays

Industry consensus: DO NOT use load in metrics
- Default EIGRP: K2=0 (load ignored)
- Default EIGRP: K4=0, K5=0 (reliability ignored)
- Only bandwidth and delay are typically used

Metric Scheme Design Example

Enterprise Campus OSPF Metric Scheme:

Goal: Prefer faster links, provide predictable failover

Reference bandwidth: 100 Gbps (100000)
router ospf 1
  auto-cost reference-bandwidth 100000

Resulting costs:
┌──────────────────────────────────────────────┐
│ Link Type           │ Speed    │ OSPF Cost  │
├──────────────────────────────────────────────┤
│ Core-to-Core        │ 100 Gbps │ 1          │
│ Core-to-Dist        │ 40 Gbps  │ 2-3        │
│ Dist-to-Access      │ 10 Gbps  │ 10         │
│ Access-to-Server    │ 1 Gbps   │ 100        │
│ Management          │ 100 Mbps │ 1000       │
│ Out-of-band backup  │ 10 Mbps  │ 10000      │
└──────────────────────────────────────────────┘

Traffic behavior:
- Normal: Fast links used (cost 1-100)
- If Core link fails: Traffic reroutes via alternate core (same cost)
- If Dist link fails: Traffic uses parallel Dist link or routes via Access
- Only extreme failure: OOB management network (cost 10000) used

Link Aggregation Considerations

Problem:
Single 10GbE: Cost 10
4x10GbE LAG: OSPF cost still 10 (just sees one interface)

But LAG provides 4x bandwidth!

Solutions:

1. Manual cost adjustment:
interface Port-channel1
  ip ospf cost 3   ! (10/4 ≈ 3, reflecting 4x capacity)
  
2. OSPF cost based on aggregate bandwidth:
   Some platforms automatically calculate LAG bandwidth
   Check: show interface port-channel1 | include BW
   
3. Accept the limitation:
   If both 10GbE and LAG are alternatives, LAG will be
   preferred due to ECMP across member links, even if
   OSPF cost is same

Routing metrics are the quantitative language routers use to compare paths. Let's consolidate what we've learned:

Module Complete

You've now completed the comprehensive exploration of Routing Concepts. You understand:

• Routing definition — What routing is and why it exists
• Routing tables — How routers store and use routing information
• Next hops — How packets move hop-by-hop toward destinations
• Default routes — The gateway of last resort for unknown destinations
• Routing metrics — How routers quantify and compare paths

These foundational concepts underpin all routing protocols and algorithms. Every decision an OSPF router makes, every BGP path selection, every EIGRP feasibility calculation builds on these fundamentals.