Comparison and Analysis of Spine Leaf Architecture and Three-Tier Architecture

Data centers have relied on traditional three-tier network architecture for several years for managing traffic. This architecture features an aggregation layer, an access layer, and a core layer. It has been majorly used to manage north-south traffic. However, with increasing dependencies on various high-speed devices, the data centers are driven to implement an architecture that can help them manage this load effectively. As a result, many are embracing leaf-spine architecture. This new networking architecture is helping data centers to effectively manage east-west traffic. This post analyzes this architecture in detail and explores why it is here to stay.

Spine Leaf Architecture

A Brief Overview of Traditional Three-Tier Architecture

As said before, this architecture comprises an access layer, a core layer, and an aggregation layer. The following pointers will help you understand these layers better.

  • Access Layer: This is the topmost layer and comprises access switches, which are positioned on the top of the rack. Due to their positioning, these switches are also known as top-of-rack switches and are used to connect servers physically.
  • Aggregation Layer: This layer comprises aggregation switches connected to the access switches. The aggregation switches can be layer 3 or layer 2 switches. It provides other services such as SSL offload, firewall, and network analysis.
  • Core Layer: This layer features a core switch, which connects multiple VLANs on layer 2. The core switch is responsible for forwarding data packets in and out of the data center. It creates a robust layer 3 network.

Limitations of Traditional Three-Tier Architecture

The three-tier architecture has certain limitations, which has paved the way for the easy adoption of spine-leaf architecture. The following pointers will help you understand it better.

  • As the demand for high-performance networks increases, businesses using traditional three-tier architecture are driven to expand their capacity. As a result, they invest in a large number of core layer and aggregation switches to meet the expanding demands. This adds to the costs of investment and network scalability requires a high level of planning.
  • The three-tier architecture is mainly designed to support north-south traffic. Although it can be used for east-west traffic, it may not perform as required. If there is a large amount of east-west traffic, there will be high pressure on the core switch and aggregation switch. When this happens, the performance of the network may jeopardize.
  • Three-tier architecture has a more complex structure than 1 or 2 tier architecture.
  • When the network expands, data centers may be distributed in different geographical locations for easy control and management. Many virtual machines (VMS) are created and migrated without changing the gateway, IP address, and other network attributes. All this would be supported only if the network has a strong layer 2. The migration of large networks becomes difficult in a three-tier architecture.

An Overview of Spine-leaf Architecture

As the name suggests, this architecture comprises a spine layer and a leaf layer. These layers are represented by spine and leaf switches, which are designed to support east-west traffic in data center. The east-west traffic is created during server-to-server communication in the same data center.

  • Leaf Layer: The leaf switches are basically layer 2 switches, which help connect storage devices and servers.
  • Spine Layer: The spine switches are layer 3 switches, which serve as the network’s backbone and help in routing and forwarding.

Advantages of Spine Leaf Architecture

The spine-leaf architecture offers several advantages to data centers, which has contributed to their easy acceptance among businesses. The following are a few advantages of employing spine-leaf architecture.

  • Scalability: As the performance demands of the network increase, the number of spine-leaf switches can be increased easily. This would help businesses optimize their scalability and meet expanding business demands easily.
  • Minimal Traffic Congestion: The spine-leaf architecture helps minimize traffic congestion by applying the shortest path bridging (SPB) and transparent interconnection of multiple links (TRILL).
  • Network Stability: The inter-layer bandwidth of the network can be increased by adding the uplinks to the spine layer of the architecture. It helps improve the network stability.
  • Improved Redundancy: The servers can be connected to the core network using a leaf switch in this type of architecture. This assures high flexibility in hyper-scale data centers. All spine switches and leaf switches can be connected to each other in this network. This would help reduce traffic bottlenecks.
  • Can Be Adapted to the Enterprise: Although the leaf-spine architecture is designed for data centers, today, it is also used for enterprise networks. The spanning tree protocol (STP) between the access layer switch and its feeder switch is removed and MLAG is employed. This MLAG enables access layer switches to connect with upstream switches for redundancy. The MLAG peer switches help reroute the traffic automatically even if a spine or a leaf switch fails, which helps ensure continuous uptime.
  • Low Costs of Implementation: Spine leaf architecture use fixed-configuration switches, which have a fixed number of ports. These switches aren’t expensive, unlike modular switches, which helps reduce the cost of implementation.
  • Low Power Consumption: Fixed configuration switches aren’t power hungry like modular switches, which helps reduce the overall power consumption of the network.

How to Design a Spine Leaf Architecture

There are many things to remember when designing and deploying spine-leaf architecture. The following are a few important ones among them.

  • Layer 2 and 3 Design: The spine leaf topology can be developed on layer 3 or layer 2. Layer 3 is for subnetting and layer 2 for VLAN configuration. In layer 3, each link will be routed and this approach works well when VLAN networks are sequestered to leaf switches that work independently. In layer 2, SPB or TRILL will replace the STP. All hosts offer a loop-free route to their MAC through a shortest path computation. These hosts are also connected to the fabric.
  • Leaf and Spine Sizing: The number of spine-leaf switches used in the topology will depend on the port density of spine switches.
  • Oversubscription Rate:When all devices in the network send traffic simultaneously, it is known as oversubscription. This traffic gets aggregated on the active link. The ratio of oversubscription must be maintained at 3:1 or less. It is measured and delineated as a ratio between downstream capacity and upstream bandwidth.

An Overview of Spine and Leaf Switches

Spine switches are the core switches, which are designed to handle the network traffic of layer 3. Each L3 port of the switch can be connected to the L2 leaf switch. Therefore, the number of leaf switches to be used in a network will depend on the number of ports on the spine switches. This, in turn, will help determine the servers to be connected to the network. Generally, firewalls, servers, or access points are not connected to these switches.
Leaf switches are access layer switches used to control the forward layer 3 and 2 traffic or traffic between the servers. The uplink ports on the leaf switches determine the spine switches to be connected. Similarly, the downlink ports of the leaf switches determine the devices to be connected.
Datacenter spine switches feature 10G/25G/40G/100G ports and various application functions, which assure quick network implementation. Datacenter leaf switches have different speed capabilities in uplink and downlink ports. For instance, the downlink ports may support 10G/25G/40G/50G/100G speeds, whereas the uplink ports may support 40G/100G speeds.
Spine and leaf switches are sometimes called scale-out architecture, as they can be added incrementally with new servers to meet the increasing demands of the network.

Spine Leaf Architecture Vs. Traditional Three-tier Architecture

Spine leaf and traditional three-tier architecture differ in many ways. The following pointers will help you understand it better.

  • Traffic Transformed: The three-tier architecture addresses north-south traffic, which is generated when client devices communicate with server devices. However, spine-leaf architecture is designed to address east-west traffic, which is generated when there is a server-to-server communication.
  • Layers: There are three layers in the three-tier architecture, including core, access, and aggregation layer. However, there are two layers in the spine-leaf architecture, where spine and leaf switches are connected to different devices.
  • Scalability: The three-tier architecture utilizes the STP protocol, which can support up to 100 switches. It may cause a blockage if there is any further expansion. However, the spine-leaf architecture allows expansion through spine switches.

Conclusion

Traditional three-tier architecture is still preferred in many data centers. However, the choice of the right architecture would depend on your expansion plans, scalability, and performance requirements. Both these network architecture require careful layout and design, as well as high-quality devices to optimize their value. Fiber optic devices are an integral aspect of these network architecture and you must source them from trusted manufacturers like VERSITRON. The company has been at the forefront of fiber optic technologies since 1960s and has successfully delivered fiber optic devices such as Fiber Optic Network Switches, high-definition transmitters and receivers, fiber optic media converters, video distribution amplifiers, and CCTV fiber optic video kits for several growing business networks and data centers.