Cloud Agnostic Data Centers - Smart Architecture Part 3

Cloud Agnostic Data Centers

The classic approach to networking typically involves building facilities with cages and rows of bare metal servers and network gear to establish connectivity. An alternative is to use Cloud Service Providers as the "New Data Center". Using a small footprint of network hardware and direct connections, organizations can build IP backbones that establish routing with cloud providers to route virtual networks.

Using a cloud agnostic approach, companies can strategically build network POPs within close proximity to cloud providers offering direct cross connects. A simple network model can be used to route traffic to and from a POP with standard Internet transit. An edge and core layer of routers is used to manage routing decisions at the POP, and an aggregation layer can be used to establish local cross connects to build routing peers. This model allows an organization to peer privately with AWS, Azure, IBM, GCE, or any other cloud provider offering similar services.

Once the network is live and peering is established, the virtual network built in the cloud is then routed using the IP backbone of the network. A virtual network in AWS East can talk to a virtual network in Azure Central all over the same private IP backbone, as if those virtual networks were actual data centers. Once this connectivity is in place, all virtual instances within the cloud provider are essentially just a server within the network.

Each cloud has networking limitations that organizations will need to creatively work around to meet their needs, but basic configurations, such as building NAT translations or using your own IP space, are simple advantages that can be implemented. This model allows for a small and cost efficient network footprint when compared to building traditional data centers as the data center burdens are consumed by the cloud provider.

Another benefit of this model is that control of the network is maintained by the organization. All routing decisions, IP space, QoS policies, etc. can be handled within the network layer which provides additional flexibility. Depending on the use case of the network, the total footprint can be as small as a single 45RU cabinet filled with only the necessary hardware needed to build the model and establish connectivity.

So What?

Over the past decade, the public Internet and cloud computing have truly gone global, with even the smallest business in the most remote part of the world leveraging the ability to spin up compute and storage with the likes of Amazon Web Services, IBM SoftLayer, Google Cloud, or Microsoft Azure. This new ubiquitous capability, coupled with the increasing focus on DevOps automation and user-friendly software applications, has deeper implications for the delivery of real-time communications over the Internet.

Quality of service and reduced costs are the main benefits to an organization’s ability to deliver diverse, resilient and redundant real-time communications associated with the adoption of these technologies. Network automation, building select peering relationships and deploying cloud-agnostic points of presence all contribute to increased quality of service. The start-up investment cost and effort for implementing these strategies has decreased dramatically in recent years and will continue to do so in the coming decades.

This post is an except of the original article  published by Channel Vision in July-August edition written by Ian Reither and Jason Craft.