Data Center Architecture Design: LAN, SAN, KVM, POD, and ToR Network Structures

Data Center Architecture Design

Data‑center wiring includes both inside‑the‑core computer room and outside‑the‑room (support space) cabling. The system architecture is divided into three distribution areas: Main Distribution Area (MDA), Horizontal Distribution Area (HDA), and Equipment Distribution Area (EDA), connected in a star‑topology from MDA to HDA to EDA.

LAN System Network Architecture

1) Adopt a three‑tier structure: core‑aggregation‑access. 2) Designate a core zone (MDA) in each computer‑room module and an aggregation zone (IDA) on each floor. 3) Each module contains two aggregation (IDA) distribution zones, with dual links between HDA and IDA. 4) Interconnect core‑to‑core and aggregation‑to‑aggregation within the same module.

SAN System Network Architecture

1) Use fiber‑direct‑attach centralized distribution. 2) Place SAN switches in the Main Distribution Area.

KVM System Network Architecture

1) Consider servers and network devices. 2) Install the KVM switch in the column‑head cabinet (HDA) and connect servers via copper cables. 3) Interconnect the KVM switch with the LAN switch.

Three Common Architectures

Distributed Architecture: In a typical distributed switching design, each server rack is linked by copper (6A/EA class 10GBase‑T) or fiber cables, with the number of cables determined by customer requirements. Auxiliary cabling is also needed for core network switches and SAN systems, covering carrier external services, main network equipment zones, high‑density zones, and SAN core/storage zones.

The distributed design uses stacked copper links between network‑equipment racks and main distribution frames, providing central cross‑connects that maximize switch and port utilization while remaining server‑agnostic and highly flexible.

POD Design Architecture

A POD is a modular, repeatable unit consisting of processing, storage, network, and application components that maximizes space modularity, scalability, and manageability in a data center.

Modular PODs allow power and cooling to be adjusted in 4 kW increments up to 30 kW per rack, enabling mixed‑density deployments and easy scaling.

ToR Switch Architecture

ToR (Top‑of‑Rack) switching provides high port density and I/O aggregation, dramatically reducing the number of cables and switches required within each POD.

A typical POD contains three rack units; the middle rack houses the ToR switch, supplying local access to all three racks. 10 GbE fiber trunks connect to the aggregation and core layers (usually in ZD or MD zones). Each server uses a converged network adapter (CNA) that carries both LAN and SAN traffic over 10 GbE links.

Within the POD, low‑cost, low‑latency 10GBase‑T copper cabling between servers and the top‑of‑rack switch reduces the number of intra‑rack cables by at least two per server, cutting adapters, transceivers, power consumption, and cooling load.

Only a few fiber links are needed to reach the aggregation layer, reducing the total number of switches, saving rack space, and lowering both capital and operational expenses.

Top‑of‑Rack (ToR) Switch Architecture

A unified I/O architecture merges Ethernet and Fibre Channel traffic onto a single 10 GbE link, simplifying rack‑level networking. Deploying access‑layer switches at the rack level reduces the required adapters, transceivers, and uplink ports. The design must support 10 GbE I/O technology at the rack level.

When a deployment requires more than 48 links per server, an additional access‑layer switch must be added to each rack to meet the higher link capacity.