Om's Brain

0. Load Balancing Algorithms

2 min read

1 Load Balancing Algorithms

explain all 3 algos ? Round Robin, Least Connections, IP Hash

⚖️ ASG

ChatGPT said: You got it, Om! Let’s break down the three main load balancing algorithms you’ll come across in Kubernetes/EKS Ingress Controllers like NGINX 👇

1️⃣ Round Robin

📌 What it does: Distributes each new request to the next server in line, cycling through all available backends (pods).

🔁 How it works:

  • 1st request → Pod A
  • 2nd request → Pod B
  • 3rd request → Pod C
  • 4th request → Pod A again … and so on.

✅ Pros:

  • Simple and fair.
  • Great when all pods have similar load capacity.

❌ Cons:

  • Doesn’t account for current load (e.g., one pod may be busier).

2️⃣ Least Connections

📌 What it does: Sends requests to the pod with the fewest active connections.

📉 How it works:

  • Tracks how many active (in-flight) requests each pod has.
  • Routes new requests to the least busy pod.

✅ Pros:

  • Better for long-running requests.
  • Dynamically balances based on real-time load.

❌ Cons:

  • Slightly more overhead (needs to track connections).
  • Not all ingress controllers support this out of the box.

3️⃣ IP Hash

📌 What it does: Uses a hash of the client’s IP address to always send requests to the same backend pod.

🔐 How it works:

  • Hash(IP) → assigns a pod.
  • All requests from that IP go to the same pod (as long as the pod is alive).

✅ Pros:

  • Helps with session persistence (sticky sessions).
  • Same user always hits the same pod — useful for caching.

❌ Cons:

  • Traffic may be uneven if certain IPs generate more load.
  • Fewer IPs = fewer hash buckets = imbalance.

🎯 Quick Comparison

AlgorithmLoad AwarenessSticky SessionsUse Case Example
Round Robin❌ No❌ NoSimple balanced traffic
Least Connections✅ Yes❌ NoAPIs with varying request durations
IP Hash❌ No✅ YesChat apps, game servers, caching apps

Mindmap

Graph View

Backlinks