Data Center Architectures

Hi everyone! In this entry we are going to talk about some different data center architectures and the strengths and weaknesses of each one of them in order to gather some useful information that will help us decide what kind of data center structure are we going to choose.

The main three different architectures that we see in actual data centers and we’ll see in the future designs are:  Distributed architecture, Fabric design and Application Driven architecture.

Distributed Data Center

This data center architecture is formed by the traditional three layer tier design (core, aggregation and access). The access layer is where all the servers are connected; usually a top of the rack or end of the row scheme is implemented to connect all of the servers to the access switches. This access switches will be interconnected by the aggregated layer switches that will forward the information to the last layer of the Data Center that is the core layer. In the core layer, a much more complex switching system will connect the aggregation layer to Internet.

Even though this is the most common architecture used in Data Centers, if you choose this kind of design you will have to face some challenges related to scalability, fault tolerance, energetic efficiency and saturation of the higher layers of the architecture.

Fabric Based Data Center

This type of design consists in a fully interconnected data center. By redundantly connecting all the racks to a switching structure formed by  spine switches we ensure a high availability system that will distribute the load of the outgoing and the internal traffic. Also the multiple path structure will provide resiliency to the network.

This architecture is fully scalable because if we need more servers we can add another rack to the spine and if we need more capacity, by adding one spine we provide a higher level of redundancy and connectivity between the existing racks.

Application Driven data Center

The applications within the data center will be the ones that control the network design depending on the processing needs and the amount of traffic generated. By automatizing the network design changes we can get an autonomous and user-friendly system. The system itself will run all the configuration changes making it easier to deploy and we will be able to add new elements to the network without problems.

This kind of design is ideal for cloud based data centers which have to handle a lot of different applications and traffic from different clients.

See you in future posts!