9 Network Optimization Strategies
This chapter covers
- Understanding and analyzing where network charges come from
- Identifying expensive patterns: cross-AZ, cross-region, and egress traffic
- Monitoring traffic with AWS CUR, Athena, CloudWatch, and VPC Flow Logs
- Using VPC endpoints, PrivateLink, and NAT alternatives for cost reduction
- Applying network optimization strategies in a multi-cloud environment
Cloud networking often hides in plain sight when it comes to cost. Unlike compute or storage, network charges typically don’t show up as standalone line items tied to a single resource; they accumulate silently across data transfers, NAT gateways, load balancers, and edge services. But left unchecked, these hidden costs can quickly erode your FinOps gains.
This chapter builds on the visibility foundation laid in Chapters 3 and 4, and the optimization techniques and automation principles from previous chapters, focusing now on how to monitor, control, and reduce data transfer and network-related costs.
9.1 Understanding Network Charges
9.1.1 Key Types of AWS Network Costs
9.2 Monitoring and Reporting AWS Network Costs
9.2.1 Getting Visibility with AWS CUR + Athena
9.2.2 Using CloudWatch to Track Network Metrics
9.2.3 Using CUDOS to Visualize Network Charges
9.3 Best Practices for Network Optimization
9.3.1 Use VPC Endpoints Instead of NAT Gateways
9.3.2 Compress Data Before Transfer
9.3.3 Enable Caching with CloudFront
9.3.4 Avoid Cross-AZ Traffic for Internal Services
9.3.5 Use VPC Flow Logs to Trace and Attribute Internal Traffic
9.3.6 Optimize Load Balancer Design
9.3.7 Tune DNS TTLs to Reduce Lookup Overhead
9.3.8 Use Direct Connect for High-Volume Hybrid Traffic
9.3.9 Use CloudWatch to Alert on Traffic Anomalies
9.4 Optimize Network Costs in Multi-Cloud Architectures
9.4.1 Route Through Peering or Direct Connections
9.4.2 Place Dependent Workloads in the Same Region
9.4.3 Compress and Batch Transfers
9.4.4 Use Object Storage as a Transfer Bridge
9.4.5 Track Cross-Cloud Traffic in CUR or Third-Party Tools
9.4.6 Forecast and Model Network Costs
9.5 Summary