Curated reference links for Network Architectures

Whether you are preparing for a network design certification, architecting your first data center fabric, or simply curious how Google and Meta build networks at planetary scale — this post collects the best reference links for six major network architecture domains, organized from vendor design guides to real-world engineering papers.

1. Enterprise Network Architecture

Enterprise (campus) networks follow the classic hierarchical model: access, distribution, and core layers, with two-tier "collapsed core" variants for smaller sites.

• Cisco Enterprise Campus Infrastructure Design Guide (PDF) — consolidated best practices across access, distribution, and core layers.
• Cisco Enterprise Campus 3.0 Architecture: Overview and Framework — design criteria, topologies, and campus services.
• CCDE Study Guide: Enterprise Campus Architecture Design (free Cisco Press chapter) — hierarchical design models, modularity, Layer 3 routing considerations, and campus virtualization.
• Cisco Community: Campus Network Design Guideline — a quick primer on three-tier vs. collapsed-core models.

2. Datacenter Network Architecture

Modern data centers have moved from the three-tier model to flattened leaf-spine (Clos) fabrics optimized for east-west traffic, typically running VXLAN/EVPN overlays.

• Spine-Leaf Network Architecture (Intelligent Visibility) — oversubscription ratios, non-blocking design, 3-stage and 5-stage Clos, super-spine scaling.
• Modern Data Center Network Architecture: Leaf-Spine & 800G Design — includes VXLAN/EVPN overlay concepts and AI-era considerations.
• Spine-Leaf vs. Traditional Data Center Architectures (STORDIS) — a clear side-by-side comparison.
• Classic academic paper: search for "A Scalable, Commodity Data Center Network Architecture" (Al-Fares et al., SIGCOMM 2008) — the foundational fat-tree/Clos paper, freely available on university sites.

3. Service Provider Network Architecture

ISP and carrier networks are built on a core / aggregation / edge model, with MPLS in the backbone, BGP policy at the edge, and peering at Internet Exchange Points.

• How the Modern ISP Network Is Actually Structured (j2sw) — a current look at PoPs, IXP peering, and multi-terabit backbone routing.
• BGP Design & Implementation: Service Provider Architecture Fundamentals — the roles of core, aggregation, and edge layers, and where BGP policy lives.
• Cisco SRND: Service Provider Network Architecture (MPLS P/PE design) — how Provider and Provider Edge routers deliver per-customer L3VPNs.
• ISP Carrier-Grade Edge Services Architecture — from the centralized core to distributed edge PoPs.

4. Cloud Provider Network Architecture

Public clouds organize infrastructure into Regions, Availability Zones, and global edge PoP networks, with software-defined constructs (VPC/VNet) on top.

• AWS Global Infrastructure (official) — plus a helpful explainer at Tutorials Dojo: AWS Global Infrastructure Cheat Sheet covering Regions, AZs, and the 700+ edge locations.
• Azure Architecture Center — Networking — hub-and-spoke, Virtual WAN, ExpressRoute, Private Link, plus AWS/GCP comparisons.
• Cloud Networking Across AWS, Azure and GCP (Alkira) — compares VPC/VNet constructs, peering, and transit models across the big three.
• AWS Networking & Content Delivery Blog — reference architectures for Transit Gateway, Cloud WAN, Direct Connect, and multicloud interconnect.

5. Hyperscaler Network Architecture

Hyperscalers (Google, Meta, AWS, Microsoft) design their own fabrics, switches, and control planes — and publish some of the most influential networking papers in the industry.

• The Evolution of Google's Jupiter Data Center Network — how Google moved beyond fixed Clos spines to an evolvable fabric.
• Jupiter Now Scales to 13 Petabits per Second — 25 years of Google data center networking milestones.
• Introducing Virgo: Google's Megascale AI Fabric — the split into a scale-up domain, an east-west RDMA accelerator fabric, and the Jupiter front-end network.
• Google Jupiter: A Decade of Evolution (deep-dive retrospective)
• Facebook's F16 Data Center Fabric (IEEE ComSoc analysis) — also search "F16 fabric" on engineering.fb.com for Meta's original post.
• Key papers to look up: "Jupiter Rising" (SIGCOMM 2015), "Jupiter Evolving" (SIGCOMM 2022), and Meta's "RDMA over Ethernet at Meta Scale" (SIGCOMM 2024).

6. Distributed Systems Network Architecture

At the application layer, networking shifts to service-to-service communication: service meshes, sidecar and ambient proxies, control/data plane separation, and eBPF-based observability.

• Distributed Systems & Microservices Architecture (open study site) — Kubernetes control plane internals, Envoy-based meshes, traffic routing, and eBPF observability.
• Service Mesh Architecture: Key Components and Design Factors (Solo.io) — sidecar proxies, data plane, and control plane explained.
• HashiCorp: Service Mesh and Microservices Networking — distributed data planes with end-to-end TLS and edge enforcement.
• Istio Official Architecture Documentation
• Academic survey: Service Mesh — Architectures, Applications, and Implementations

How to Use These References

The vendor design guides (Cisco, Azure) give you the textbook layered models, while the Google and Meta engineering blogs and SIGCOMM papers show how those models break down at extreme scale — and what replaces them. Reading one of each side by side (for example, the Cisco campus guide next to Jupiter Rising) makes the architectural trade-offs much clearer than either alone.

Found this useful? Leave a comment below with the architecture domain you'd like a deep-dive on next.