TechPrufer
Get started
Backblaze
Backblaze logoBackblaze · Feature trail

Network Traffic Analysis and Capacity Planning

4 posts · Jun 2026 – Jul 2026

What this trail covers

Backblaze continues to analyze network traffic patterns, with a focus on the impact of AI workloads. This quarter's report introduces variance analysis to distinguish stable traffic from volatile AI-driven patterns. The analysis covers total traffic volume, magnitude (bits per IP), and communication uniqueness across different network types (CDN, Hosting, Hyperscaler, ISP-regional, Neocloud) and regions (US-West, US-East, CA-East, EU-Central). The data highlights the dynamic nature of neocloud and AI-driven traffic, emphasizing the need for flexible infrastructure that can adapt to evolving AI ecosystems and the importance of observing network telemetry over time to distinguish durable trends from short-term experimentation.

All Backblaze trails
  1. Network Stats for Q2 2026: All Eyes on Neocloud Traffic VarianceThis post introduces variance analysis to the network traffic reporting, using statistical baselines and SciPy to distinguish stable traffic from volatile AI workloads. It presents new charts and heatmaps focusing on variance, magnitude, and uniqueness for neocloud and hyperscaler traffic, contrasting them with traditional traffic patterns. The analysis identifies increased neocloud and ISP-regional traffic in Q1/Q2 2026, a new baseline for CDN traffic, and growing magnitude in EU-Central neocloud traffic.
  2. Why AI Projects Stall: Data SilosThis post identifies data silos as a major impediment to AI project success. It advocates for centralizing multimodal data, establishing data governance early, and planning storage infrastructure with long-term costs and retrieval latencies in mind. The post emphasizes that proprietary data is the key differentiator for AI initiatives and that aligning AI strategy with data strategy leads to better infrastructure decisions.
  3. What Network Data Can and Can’t Tell Us About AI InfrastructureThis post analyzes network telemetry data to understand the characteristics of AI infrastructure traffic. It details how network volume, connection persistence, endpoint concentration, and geographic patterns reveal operational demands. The post highlights that while network data shows current AI workload behavior (e.g., elephant flows, regional clustering), it cannot definitively predict future infrastructure design due to the rapid evolution of AI technologies and compute availability. It emphasizes that interpreting network telemetry requires observing trends over multiple quarters and that flexible, interoperable tech stacks are crucial for adapting to the dynamic AI ecosystem.