1.7 Datacentre network architecture

1.7 Explain basic corporate and datacentre network architecture.

• Three-Tiered Network Architecture

  • A traditional corporate network architecture has three layers, each serving a specific function:
    • Core
      • The core layer is the network backbone, providing high-speed, reliable connectivity across the organization. It connects major parts of the network, like data centres and branch offices, ensuring efficient data transfer. The focus here is on low latency, high bandwidth, and high availability.
    • Distribution/Aggregation Layer
      • This layer aggregates data from the access layer and routes it to the core. It often handles network policies (such as security, routing, and quality of service) and inter-VLAN routing. The distribution layer acts as a boundary between the core and access layers and can include features like redundancy and load balancing.
    • Access/Edge Layer
      • The access layer connects end devices (computers, IP phones, printers) to the network. It manages the direct connection of devices to the network and is often responsible for applying network policies close to the user. Access switches and wireless access points are commonly used at this level.

• Software-Defined Networking (SDN)

  • Software-defined networking is an approach to network management that enables centralized, programmatically configured control of the network using software.
    • Application Layer
      • This is where applications that require network resources interact with the network. The application layer makes requests for connectivity, bandwidth, or other network services based on application needs.
    • Control Layer
      • The control layer is the “brain” of SDN, managing and directing how data packets move through the network. It communicates with both the infrastructure layer and application layer, enforcing policies and making real-time routing decisions.
    • Infrastructure Layer
      • This layer includes the physical network equipment like switches and routers that handle the data. The infrastructure layer implements the instructions provided by the control layer to manage data flows according to network policies.
    • Management Plane
      • The management plane is used to configure, monitor, and manage the network. It provides the interface to control and monitor the other layers, often allowing administrators to define policies and monitor network health.

• Spine and Leaf Architecture

  • A modern data centre network topology, commonly used in SDN environments, that provides a scalable and efficient way to handle high amounts of East-West traffic
    • Software-Defined Network (SDN)
      • SDN is often integrated into the spine-leaf architecture, where centralized control software manages data flows and can dynamically adjust to changing network demands.
    • Top-of-Rack (ToR) Switching
      • In this setup, a switch is located at the top of each server rack (leaf), connecting the servers in that rack to the rest of the network. ToR switching minimizes cable runs and improves efficiency within each rack.
    • Backbone
      • In a spine-leaf topology, the spine switches act as the backbone. They interconnect all the leaf switches to provide high-speed, non-blocking connections between racks. The spine layer can quickly route data between different parts of the data center.

• Traffic Flows

  • Traffic in data centres is categorized into two main types based on its direction and purpose:
    • North-South Traffic
      • This traffic flows between an internal network and an external network, like from a data centre to the internet or between branch offices and headquarters. It’s often associated with client-to-server communications.
    • East-West Traffic
      • East-West traffic moves laterally within a data centre or internal network, often between servers or virtual machines. It’s common in applications where services are distributed across multiple servers, like microservices-based applications.

• Branch Office vs. On-Premises Data Centre vs. Colocation

  • These terms refer to the physical and operational locations where corporate data and applications are hosted:
    • Branch Office
      • A branch office is a remote location separate from the company’s main headquarters or data centre. Branch offices typically connect to the central network via VPN or direct WAN links and may rely on central resources for computing power.
    • On-Premises Data Centre
      • An on-premises data centre is a facility owned and operated by a company on its own premises. It houses servers, storage, and networking equipment within a controlled environment and provides full control over physical and operational security.
    • Colocation
      • A colocation data centre is a third-party facility where a company can lease space to house its servers and other IT equipment. Colocations offer power, cooling, and connectivity, while the organization retains control over its hardware and data.

• Storage Area Networks (SAN)

  • A SAN is a high-speed network that provides access to consolidated, block-level storage, which is typically used by data centres.
    • Connection Types
      • Fibre Channel over Ethernet (FCoE)
        • This protocol allows Fibre Channel storage traffic to be transmitted over Ethernet networks, reducing the need for separate infrastructure for storage and networking.
      • Fibre Channel
        • A high-speed network protocol primarily used for SANs, providing dedicated, high-throughput connections for storage traffic.
      • Internet Small Computer Systems Interface (iSCSI)
        • iSCSI allows SCSI commands to be sent over IP networks, making it possible to use standard Ethernet for SAN connections without specialized hardware.

---

Links to Videos and Additional reading

Professor Messer -1.7 – Network Architecture

• Network Architectures (9:54)
• Storage Area Networks (3:20)