N10-009 Question 107
Single answerMesh, Hybrid, Star/hub and spoke, Spine and leaf, Point to pointA company is building a new virtualized data center to host hundreds of east-west traffic flows between application servers, databases, and storage nodes. The network team wants a design that provides predictable low latency between any two racks and can scale horizontally as more racks are added. Traditional three-tier switching has already caused bottlenecks in a pilot environment. Which topology is the BEST choice for this deployment?
- A
Star (hub-and-spoke) topology with a single core switch pair handling all rack uplinks
- B
Full mesh topology between every top-of-rack switch
- C
Spine-and-leaf topology with each leaf switch uplinked to every spine switch
- D
Point-to-point topology with dedicated links only between the most critical server pairs
Show answer and explanation
Correct answer: C
Explanation
The best answer is the spine-and-leaf topology because it addresses the core requirements in the scenario: high east-west traffic, low and predictable latency, and horizontal scalability. In a spine-and-leaf fabric, every leaf switch has uplinks to every spine switch, so traffic between devices in different racks typically traverses a consistent path pattern with predictable performance. This model is widely used in modern data centers because traditional hierarchical topologies can create bottlenecks when server-to-server traffic becomes more common than north-south client-to-server traffic. Full mesh may seem attractive for performance, but it becomes operationally unmanageable as the environment grows. Star or hub-and-spoke designs are simpler but create aggregation chokepoints. Point-to-point connections solve only isolated communication needs, not fabric-wide connectivity. These design principles align with standard networking best practices for scalable data center architectures and are consistent with how modern vendors document leaf-spine deployments for east-west optimized environments.
- A. Incorrect.
This is incorrect. A star or hub-and-spoke design centralizes traffic through core devices, which can work well in many enterprise networks and WANs, but it is less ideal for a modern data center with heavy east-west traffic. As the number of rack-to-rack flows increases, the core becomes a concentration point and can introduce oversubscription and latency bottlenecks.
- B. Incorrect.
This is incorrect. A full mesh between every top-of-rack switch would provide high path redundancy and low hop counts, but it does not scale well. The number of links and ports grows rapidly as racks are added, making cabling, switch port consumption, and management impractical in real-world data centers.
- C. Correct.
This is correct. Spine-and-leaf is specifically designed for scalable data center fabrics with high east-west traffic. Each leaf connects to every spine, creating predictable latency and consistent hop counts between racks. The design also scales horizontally by adding additional leaf switches for more racks and spine switches for more fabric capacity, which aligns with common modern data center best practices.
- D. Incorrect.
This is incorrect. Point-to-point links are appropriate when connecting two specific devices or locations directly, such as a WAN circuit or dedicated interconnect. They are not an efficient topology for a large virtualized data center where many racks must communicate with many other racks dynamically.