Problem
Default round-robin ignores load, latency, and session needs. Wrong algorithm = uneven load, high latency, broken sessions.
distribution-communicationestablished · medium operational burden
Load Balancing Algorithms
Also known as: lb-algorithms, round-robin, least-connections, consistent-hashing-lb, weighted
Intent
Choose the right load balancing algorithm: round-robin, least connections, least latency, consistent hashing — trade-offs for latency, distribution, and session affinity.
Forces
- •Distribution: even vs weighted vs locality-aware
- •Latency: route to fastest vs simplest algorithm
- •Session affinity: sticky sessions vs stateless
- •Health awareness: avoid unhealthy instances
Solution
✓ When to Use
- •Any service with multiple instances
- •Heterogeneous fleet (different instance types)
- •Session affinity needed (legacy apps)
- •Latency-sensitive user-facing services
✗ When Not to Use
- •Single instance
- •Stateless, homogeneous, low scale (round-robin fine)
Pros
- +Right algorithm = even load, low latency, high availability
- +Health awareness = automatic failover
- +Session affinity = legacy app support
Cons
- −Complexity: algorithm choice, tuning, monitoring
- −Session affinity = uneven load, failover complexity
- −Latency-aware: oscillation, noise sensitivity
Cost Profile
Infrastructure
Low (software LB) to High (hardware, global)
Operational
Medium — algorithm tuning, health checks
Cognitive
Low — standard algorithms
Failure Modes
Unhealthy instance in pool: errors, latency
Sticky session imbalance: one instance overloaded
Hash skew: consistent hash → hot instance
Health check flapping: instance flaps in/out
Split-brain: two LBs, different views
Real-World Examples
Alternatives
- service-discovery
- round-robin
- least-connections
- consistent-hashing
- random
Related Patterns
- service-discovery
- health-checks
- circuit-breaker
- consistent-hashing
- session-affinity
- kubernetes-service
Competency Domains
distribution communicationreliability opsscalingeconomics evolutiondeployment