Serverless Function-As-A-Service (FaaS) Platforms

Traditional cloud computing services such as Infrastructure-as-a-Service (IaaS), Container-as-a-Service (CaaS) offer virtualized resources such as Virtual Machines (VMs) or containers as services to the user. The principle idea in the virtualized service off erings is the concept of a `server’ machine. Servers have been used extensively to build and host applications, however, servers are difficult to setup, confi gure, monitor and manage for operational use. It often requires specialized personnel for management, and its start-up times are slow, thereby restricting its ability to quickly scale.While IaaS provides complete choice and control of the resources, the user is burdened with the responsibility to select deployment platforms and strategies as- sociated with hosting the application. In a shift towards providing ease-of-use and reduced resource management to the user, serverless computing o ffers much higher abstraction where the cloud provider takes the responsibility of selecting and managing the resources. Function as a Service(FaaS) is one of the recent popular cloud computing service o fferings. It is a form of serverless computing where the applications are composed as stateless functions and are event driven. FaaS abstracts the provisioning and administration of servers, automatically performs the deployment of the application, and lets the user focus on the development of the application’s business logic. FaaS diff ers from the serverless computing because of the agility of the deployment of functions, coupled with powerful sub-second billing method. The billing model of FaaS computing platform is more fine-grained compared to the traditional cloud computing such as containers and VMs which are billed by the hour or minutes. The `pay-as-you-use’ model bills the user based on the amount of time the function is executed and the size of function memory that was used for it and doesn’t bill for idle time, this is a radical shift from traditional billing models where the pricing is charged for the amount of time a resource is leased even if the resource was unused or idle.

Ongoing research:

    1. QoS Aware FaaS Platforms

    2. Isolates as constructs for FaaS