As defined elsewhere in this book, reliability is resiliency plus elasticity. That is, IT infrastructures ought to be extremely elastic. They have to be application-aware, and to meet up any kind of spikes in user and data loads, infrastructure modules and assets have to be elastic. The popular compute instances such as virtual machines and containers have to be automatically provisioned in addition to the existing ones to tackle the extra load. Similarly, other infrastructural components such as networking solutions and storage appliances have to be enabled to be auto-scaling in times of need. These are coarse-grained IT resources. The fine-grained IT resources such as memory, processing power, and I/O capability also have to have the power to be self-scaling in an on-demand manner. Thus, any spike in load can be met by IT resources in an automated manner with less intervention, involvement, and interpretation of human resources. In the cloud era, additional IT modules can be provisioned across nearby availability zones by taking the location constraints into account. The intelligent capacity planning and management acquires special significance here. Not only infrastructures, but also applications have to be architected and designed in such a manner to support auto-scaling intrinsically. There are patterns, procedures, and practices aplenty to come out with highly scalable applications and services. With the deployment of web-scale applications and as the traffic varies very frequently in a big way, the auto-scaling feature is being insisted these days.

Real-time scalability: Provisioning additional resources to meet up increasing demands is being sped up through the leverage of application containers. Containers, as articulated in other chapters, are lightweight and hence bringing forth additional containers concurrently is faster and easier. Thus, the goal of real-time scalability facility is being realized through the containerization movement. Containers typically take a couple of seconds to be alive whereas virtual machines consumer a few minutes. Bare metal servers take several minutes to be ready for receiving client's requests. Thus, considering the limitations of physical and virtual machines, horizontal scalability gains prominence these days.