Routing policies (simple, weighted, latency) in AWS - Deep Dive

Overview - Routing policies (simple, weighted, latency)

What is it?

Routing policies in AWS are rules that decide how internet traffic is directed to different servers or resources. Simple routing sends all traffic to one place. Weighted routing splits traffic based on assigned weights, like sharing slices of a pie. Latency routing sends traffic to the server that responds fastest to the user.

Why it matters

Without routing policies, all users might be sent to the same server, causing slow responses or crashes when traffic is high. Routing policies help balance the load, improve speed, and increase reliability, making websites and apps work smoothly for everyone.

Where it fits

Before learning routing policies, you should understand basic DNS concepts and how domain names connect to IP addresses. After mastering routing policies, you can explore advanced routing like failover and geolocation routing to optimize global traffic.

Mental Model

Core Idea

Routing policies control how user requests are shared or directed among multiple servers to optimize performance and reliability.

Think of it like...

Imagine a restaurant with multiple chefs. Simple routing is like sending all orders to one chef. Weighted routing is like giving some chefs more orders than others based on their skill or speed. Latency routing is like sending orders to the chef who can cook fastest right now.

┌───────────────┐
│ User Request  │
└──────┬────────┘
       │
       ▼
┌───────────────┐
│ Routing Policy│
├───────────────┤
│ Simple        │
│ Weighted      │
│ Latency       │
└──────┬────────┘
       │
       ▼
┌───────────────┬───────────────┬───────────────┐
│ Server A      │ Server B      │ Server C      │
│ (All traffic) │ (Weighted %)  │ (Lowest delay)│
└───────────────┴───────────────┴───────────────┘

Build-Up - 6 Steps

1

Foundation

Understanding DNS and Traffic Direction

Concept: Learn how DNS translates domain names to IP addresses and directs user traffic.

DNS (Domain Name System) is like the internet's phone book. When you type a website name, DNS finds the IP address of the server hosting it. This process directs your request to the right place.

Result

User requests reach the correct server based on DNS lookup.

Understanding DNS basics is essential because routing policies build on how DNS directs traffic.

2

Foundation

What is Simple Routing Policy?

Concept: Simple routing sends all traffic to a single resource without splitting or conditions.

In AWS Route 53, simple routing means all users go to one server or IP address. It's like having one entrance to a store where everyone lines up.

Result

All user requests go to the same server.

Knowing simple routing helps grasp how routing policies can be more complex to improve traffic handling.

3

Intermediate

Weighted Routing Policy Basics

🤔Before reading on: do you think weighted routing sends equal traffic to all servers or splits it unevenly? Commit to your answer.

Concept: Weighted routing splits traffic among multiple resources based on assigned weights.

You assign numbers (weights) to servers. Higher weight means more traffic. For example, if Server A has weight 70 and Server B has 30, about 70% of users go to A and 30% to B.

Result

Traffic is divided proportionally among servers based on weights.

Understanding weighted routing reveals how traffic can be balanced to optimize resource use and test new servers.

4

Intermediate

Latency Routing Policy Explained

🤔Before reading on: do you think latency routing sends traffic randomly or to the fastest server? Commit to your answer.

Concept: Latency routing sends user requests to the server with the lowest network delay.

AWS measures the time it takes for a server to respond to a user. The routing policy sends the user to the server that responds fastest, improving speed and experience.

Result

Users connect to the server with the lowest latency, reducing wait times.

Knowing latency routing helps improve user experience by minimizing delays based on geography and network conditions.

5

Advanced

Combining Routing Policies for Resilience

🤔Before reading on: can you combine weighted and latency routing in AWS Route 53? Commit to your answer.

Concept: AWS allows combining routing policies to balance load and optimize speed simultaneously.

You can create weighted records within latency-based routing. For example, among low-latency servers, traffic can be split by weight to test new versions or balance load.

Result

Traffic is routed to the fastest servers and split by weight for fine control.

Understanding policy combinations enables sophisticated traffic management for high availability and performance.

6

Expert

Limitations and Edge Cases of Routing Policies

🤔Before reading on: do you think latency routing always picks the fastest server regardless of server health? Commit to your answer.

Concept: Routing policies depend on health checks and can behave unexpectedly if health data is stale or missing.

If a server is down but health checks fail to detect it, latency routing might still send traffic there, causing errors. Weighted routing without health checks can overload unhealthy servers. Proper health checks and failover policies are critical.

Result

Traffic may be misrouted if health checks are not configured properly.

Knowing routing policy limits prevents downtime and ensures traffic goes only to healthy servers.

Under the Hood

AWS Route 53 uses DNS queries to direct traffic. When a user requests a domain, Route 53 evaluates the routing policy rules and health checks to decide which IP address to return. For weighted routing, it uses randomization weighted by assigned values. For latency routing, it uses latency measurements from AWS edge locations to the servers. The DNS response changes dynamically based on these rules.

Why designed this way?

Routing policies were designed to improve user experience and resource utilization by controlling traffic flow. Simple routing is easy but limited. Weighted routing allows gradual rollouts and load balancing. Latency routing optimizes speed by considering network delays. These policies evolved to meet diverse needs of global applications.

User Request
   │
   ▼
┌───────────────┐
│ Route 53 DNS  │
│ Resolver      │
└──────┬────────┘
       │
       ▼
┌───────────────┐
│ Evaluate      │
│ Routing Policy│
│ (Simple,      │
│ Weighted,     │
│ Latency)      │
└──────┬────────┘
       │
       ▼
┌───────────────┬───────────────┬───────────────┐
│ Return IP of  │ Return IP of  │ Return IP of  │
│ Server A     │ Server B      │ Server C      │
└───────────────┴───────────────┴───────────────┘

Myth Busters - 4 Common Misconceptions

Quick: Does weighted routing guarantee exact traffic percentages every second? Commit to yes or no.

Common Belief: Weighted routing sends exactly the assigned percentage of traffic all the time.

Reality: Weighted routing approximates traffic distribution over time but individual DNS queries may not reflect exact weights instantly.

Why it matters: Expecting precise traffic splits can lead to confusion during testing or rollout, causing misinterpretation of results.

Quick: Does latency routing always pick the physically closest server? Commit to yes or no.

Common Belief: Latency routing sends traffic to the geographically closest server.

Reality: Latency routing chooses the server with the lowest network delay, which may not be the closest geographically due to internet routing paths.

Why it matters: Assuming geographic proximity equals lowest latency can cause wrong server placement decisions.

Quick: If a server is down, will routing policies automatically stop sending traffic there? Commit to yes or no.

Common Belief: Routing policies automatically avoid unhealthy servers without extra setup.

Reality: Routing policies rely on health checks configured by the user; without them, traffic may still be sent to down servers.

Why it matters: Not configuring health checks can cause downtime and poor user experience.

Quick: Can you combine simple, weighted, and latency routing policies in one DNS record? Commit to yes or no.

Common Belief: All routing policies can be combined freely in one DNS record.

Reality: AWS Route 53 restricts combining certain routing policies; for example, simple routing cannot be combined with weighted or latency routing in the same record set.

Why it matters: Misunderstanding policy combinations can cause configuration errors and unexpected traffic behavior.

Expert Zone

1

Latency routing depends on AWS edge locations' latency measurements, which may vary over time and affect routing decisions dynamically.

2

Weighted routing can be used for gradual deployment of new application versions by shifting traffic slowly from old to new servers.

3

Health checks must be carefully designed to avoid false positives or negatives, which can cause traffic to be sent to unhealthy or overloaded servers.

When NOT to use

Avoid latency routing when your application requires strict geographic compliance or data residency, where geolocation routing is better. Weighted routing is not suitable for real-time load balancing; use load balancers instead. Simple routing is only good for single resource setups without failover needs.

Production Patterns

In production, teams use weighted routing for blue-green deployments to test new versions with a subset of users. Latency routing is common for global applications to reduce user wait times. Combining latency with weighted routing allows balancing load among fastest servers. Health checks and failover routing are paired with these policies for high availability.

Connections

Load Balancing

Routing policies complement load balancers by directing traffic at the DNS level before it reaches servers.

Understanding routing policies helps grasp how traffic is distributed globally before local load balancers handle it within data centers.

Content Delivery Networks (CDNs)

Latency routing shares the goal of CDNs to reduce delay by serving content from the closest or fastest location.

Knowing latency routing deepens understanding of how CDNs optimize user experience by minimizing network delays.

Traffic Engineering in Transportation

Routing policies are like traffic signals and route choices that direct vehicles to reduce congestion and travel time.

Recognizing this connection shows how principles of balancing load and minimizing delay apply across networks and roads.

Common Pitfalls

#1 Not configuring health checks with routing policies.

Wrong approach:

Create weighted routing records without attaching health checks, assuming traffic will avoid down servers.

Correct approach:

Attach health checks to routing records so Route 53 can detect unhealthy servers and stop sending traffic to them.

Root cause:

Misunderstanding that routing policies alone do not monitor server health; health checks are a separate but necessary feature.

#2 Assigning weights that do not sum logically.

Wrong approach:

Set weights like 100 for Server A and 1 for Server B expecting 50/50 traffic split.

Correct approach:

Assign weights proportional to desired traffic, e.g., 50 and 50 for equal split or 70 and 30 for 70%/30%.

Root cause:

Confusing weight values as percentages rather than relative proportions.

#3 Expecting latency routing to always pick the closest server geographically.

Wrong approach:

Deploy servers based only on geographic proximity assuming latency routing will pick the nearest one.

Correct approach:

Measure actual latency and consider network paths, not just geography, when placing servers.

Root cause:

Assuming physical distance equals network speed, ignoring internet routing complexities.

Key Takeaways

Routing policies in AWS control how user requests are directed to servers to improve performance and reliability.

Simple routing sends all traffic to one resource, weighted routing splits traffic by assigned proportions, and latency routing sends traffic to the fastest responding server.

Health checks are essential to ensure routing policies send traffic only to healthy servers and avoid downtime.

Combining routing policies and understanding their limits allows sophisticated traffic management for global applications.

Misconceptions about exact traffic splits, geographic proximity, and automatic health detection can cause serious production issues.