How cloud and networks achieve 99.999% availability in different ways
As we move from classic network-to-cloud-based technologies, some fundamental beliefs will be challenged. A network availability of 99.999 percent or "five 9s" has represented the desired DNA of a network. But with the introduction of cloud technologies, the way to reach those five 9s will be challenged.
The network model
Achieving 99.999 percent availability in networks is based on a well-proven recipe. We build networks from carrier-grade components. We use product architectures that have redundancy at the blade-, fan-, and power unit–levels. We test the network and software appliances rigorously. Finally, we deploy the network components and software in networks that have 1+1 and N+1 redundancy and have secure support agreements that cover parts nearby. The price tags—in terms of lead times, utilization rates, and support needs—are also very clear.
The datacenter model
In a datacenter, high availability is typically achieved in a different way. Datacenters use hardware that does not have carrier-grade requirements. The failure rates of this hardware are well known and taken into consideration when designing datacenter infrastructure. Then, as time goes on, new hardware and software are continuously deployed into the datacenter. The software is designed to avoid certain types of outages, while hardware components are replaced at frequent intervals rather than immediately after a failure.
Bridging the two worlds
A major conundrum in deploying cloud infrastructure for network purposes is to bridge the gaps between the two worlds. It is not obvious to a network professional that the cloud model can deliver 99.999 percent availability. In the same way, a cloud professional might think the cost of providing the required network availability would be prohibitively high.
So it is essential that we find a recipe for high availability within cloud-supporting network functions. It is unlikely to be either of the two models but, instead, a blend. And beyond resiliency design, we now have access to software-defined infrastructure as a new tool in the toolbox.
Predictions for the future
- 99,999 percent availability will remain the norm as we move to 5G, IoT, and cloud. The importance of service availability is increasing rather than decreasing as whole societies become dependent on cloud and mobile devices.
- The model that achieves high availability within cloud will be a blend of the two very different paradigms that are used for networks and classic IT datacenters.
- An essential part of bridging the gap is to ensure that the expertise of both network and cloud professionals is used in creating the new model.