Gaffer: Cloud Computing based Serverless Orchestration Framework for Unprecedented Workflow

Abstract

Modern serverless orchestrations are based on either simple step chaining or function invocations. This raises the challenges of implementing high-performance systems. In the context of this, the paper explores how a system can be structured to scale across cross-platform infrastructures by design. This research work has proposed a framework that can be used under circumstances where heavy computation with data consistent results are demanded from the serverless networks, in which the instances depend upon multiple nested servers and the flow of the data is unprecedented without worrying about concurrency and function invocation limitation resulting in zero throttling of requests on the service. This means that the function that needs to be triggered next will be decided by the present function which can demand multiple arguments to run that are to be provided by some other functions. The algorithm is based on the Gaffer Framework which is designed to take care of resource allocation, maintaining the function(servers) level order dependency also sub-level dependency, and resource integration using event-driven semaphore signals. The framework proposed in this paper demonstrates high coordination-free consistency along with the performance that exceeds the current orchestrations. Gaffer Framework run-time system also takes care of distributed data partitioning and efficient handling of machine failures. This makes it simple for programmers with little or no expertise with serverless and distributed systems to create and maintain large interconnected serverless containers.