data center architecture
The classic architecture ofdata centers is divided into three layers that form a hierarchical tree structure: Core Layer, Aggregation Layer and Access Layer.
The lower level of the data center architecture is the access layer or edge layer with its layer 2 switches or access switches. The servers, server farms and host computers are connected to these. The access switches are usually located in the upper part of the racks and are therefore also referred to as ToR switches.
The layer above is the aggregation layer or distribution layer, which forms the service layer of the data center. It uses the Spanning Tree Protocol( STP) to establish redundant connections to the components of the Access Layer via the Layer 2/3 switches. The aggregation layer switches provide load balancing, firewall protection, intrusion detection, and network analysis, among other functions. They are also used as connection points for uplinking to the core layer via 10 Gigabit Ethernet( 10GbE). However, 10GbE downlinks to the access layer can also be implemented in order to provide the access layer with high bandwidth.
The core layer forms the top of the tree topology and is the center of the data center network. It provides routing between other parts of the data center and outsourced data centers via the Layer 3 routers, as well as the connection to the switches on the aggregation layer and to the Internet. For this reason, extremely powerful routers with low latency are used in the core layer, via which high-speed connections are switched to the aggregation layer switches via 10 Gigabit Ethernet.
With the model shown, a balanced distribution of the data streams is achieved. However, bottlenecks can occur due to the oversubscription of uplinks between layers, resulting from the flow of data between layers and the blocking of redundant links. The leaf-spine architecture eliminates this disadvantage.