[Paper Review] Cloud Programming Simplified: A Berkeley View on Serverless Computing

Eric Jonas, Johann Schleier-Smith+12 more|arXiv (Cornell University)|Feb 9, 2019

Cloud Computing and Resource Management24 references420 citations

TL;DR

This paper presents a comprehensive vision for serverless computing as the future of cloud programming, advocating for a paradigm shift akin to the move from assembly to high-level languages. It proposes that serverless computing—comprising FaaS and BaaS—will simplify cloud development by eliminating low-level system management, and identifies key research challenges in storage, coordination, performance, security, and hardware to unlock its full potential.

ABSTRACT

Serverless cloud computing handles virtually all the system administration operations needed to make it easier for programmers to use the cloud. It provides an interface that greatly simplifies cloud programming, and represents an evolution that parallels the transition from assembly language to high-level programming languages. This paper gives a quick history of cloud computing, including an accounting of the predictions of the 2009 Berkeley View of Cloud Computing paper, explains the motivation for serverless computing, describes applications that stretch the current limits of serverless, and then lists obstacles and research opportunities required for serverless computing to fulfill its full potential. Just as the 2009 paper identified challenges for the cloud and predicted they would be addressed and that cloud use would accelerate, we predict these issues are solvable and that serverless computing will grow to dominate the future of cloud computing.

Motivation & Objective

• To advocate for serverless computing as the next evolution in cloud programming, simplifying development by abstracting away infrastructure management.
• To identify and analyze critical limitations of current serverless platforms, particularly in storage, coordination, performance, and security.
• To outline a research agenda for overcoming these limitations and enabling serverless computing to support a broader range of applications.
• To predict that serverless will become the default paradigm in cloud computing, replacing traditional serverful models.

Proposed method

• Proposes a unified definition of serverless computing as FaaS + BaaS, emphasizing automatic scaling and usage-based billing.
• Analyzes five real-world applications (e.g., ExCamera, Numpywren, Cirrus) to demonstrate current platform limitations.
• Identifies key technical challenges across abstraction, system design, networking, security, and computer architecture.
• Introduces the concept of 'serverless ephemeral' and 'durable' storage to address performance and persistence needs.
• Proposes that high-level programming abstractions and fine-grained isolation in serverless can enhance security and simplify secure programming.
• Uses historical parallels to the 2009 Berkeley View paper to argue that current challenges are solvable and will lead to widespread adoption.

Experimental results

Research questions

• RQ1 How can serverless computing overcome current limitations in storage performance and durability for fine-grained operations?
• RQ2 What system-level mechanisms are needed to enable efficient, low-latency coordination across serverless functions?
• RQ3 How can serverless platforms achieve predictable and low-latency performance for standard communication and data access patterns?
• RQ4 What architectural and security mechanisms can be leveraged to harden serverless systems against side-channel attacks?
• RQ5 How can future hardware heterogeneity and specialized processors be integrated into serverless computing to improve cost-performance?

Key findings

• Serverless computing is predicted to become the dominant paradigm in cloud computing, replacing traditional serverful models.
• Current platforms face significant limitations in fine-grained storage, coordination, and predictable performance, especially for stateful and data-intensive workloads.
• Ephemeral and durable serverless storage systems are needed to support high-performance, low-latency applications.
• Security is enhanced in serverless due to fine-grained isolation and reduced attack surface, though side-channel risks remain.
• The shift to serverless will enable lower costs and higher developer productivity, with billing models evolving to make serverless cost-competitive or cheaper than serverful computing.
• Hybrid and on-premises deployments will decline over time, though regulatory constraints may preserve some legacy use cases.