Realizing the Vision of Sky Computing
Computing can be a utility, like telephone service, unlike today's cloud markets, which introduce vendor lock-in and aggressive pricing.
Introduction
Stoica and Shenker proposed the concept of sky computing and its potential to transform the cloud computing paradigm [1]. At its core, they revisit McCarthy's vision of computing as a utility, analogous to telephone systems. In utilities, users typically do not concern themselves with the specific provider of connectivity. For instance, in telephone systems, multiple companies compete to offer the best overall service. Although they all provide the same core service—a phone number and the ability to call anyone—they differentiate through competitive pricing, add-ons, and quality of service. This competition drives higher-quality services for customers at lower costs. However, when it comes to computing, this is not the case. Major cloud computing companies dominate the execution of most web workloads with proprietary services and APIs that lock in customers, making it practically impossible to switch providers.
Stoica and Shenker argue that creating a layer of interconnectivity standards and agreements between cloud providers could foster competition and provide developers with a cloud-agnostic standard for building and deploying applications. They also suggest that smaller clouds have the greatest incentive to initially adopt such standards, as they seek to expand their service offerings beyond their native capabilities. Unfortunately, several years after the inception of this idea, progress toward the vision of sky computing has been minimal. The high cost of compute resources, especially for AI/ML workloads, only underscores the urgent need for a solution, as these costs increasingly exclude smaller players from the market.
In this manuscript, we argue that blockchain technology can serve as the key enabler for realizing the vision of sky computing by:
Providing a permissionless and open compute plane to power interconnectivity standards.
Introducing financial instruments (e.g., payments, auctions, loans) to facilitate and streamline inter-cloud agreements.
Leveraging an ecosystem of pre-existing ``clouds,'' where blockchains offer abundant compute resources for applications, creating a competitive environment of smaller vendors eager to adopt interconnectivity frameworks.
Permissionless Standard for Interconnectivity.
While many standalone products (like databases or deployment tools) support multi-cloud instantiations, services such as naming, resource discovery, and scheduling need to operate globally across multiple clouds. Blockchain technology provides a unique platform to host these services and match supply (compute resources) with demand (developers). Smart contracts can be designed to handle naming, resource discovery, and resource allocation, enabling open access for anyone to post updates or query the system at minimal cost. Developers worldwide can propose governance upgrades, and cloud providers can contribute their resources while participating in rule creation.
Blockchains ensure fair and transparent scheduling policies that prioritize user constraints and align with the network’s overarching ideology. Users can trust that workloads are scheduled in accordance with their specified requirements. The blockchain’s open and decentralized nature eliminates the need to designate a single cloud provider to operate these services, thus avoiding vendor lock-in. By offering a resilient, impartial environment, blockchain technology addresses these challenges effectively.
Financial Instruments for Inter-Cloud Agreements.
As noted in [1], facilitating agreements between cloud providers is crucial to reducing costs for users, particularly for data transfers, which are currently among the most expensive cloud operations. Blockchains can codify these agreements and provide cost-effective payment rails, enabling transactions at fractions of a cent compared to traditional systems that charge 2–3% in processing fees. These agreements, encoded in smart contracts, are transparent and auditable, with payments settled globally within seconds.
Cloud providers often operate across multiple jurisdictions and borders, making blockchain-based payment systems the most efficient way to settle agreements. Furthermore, blockchain technology enables more sophisticated price discovery mechanisms. For example, a user could auction a workload (e.g., for AI training) to the cloud provider offering the best price, with the auction logic encoded in a smart contract. Additionally, cloud providers could tokenize their compute resources and offer them as credits to startups in exchange for equity, fostering innovation and collaboration.
Bootstrapping the Sky Computing Ecosystem.
One of the most significant challenges in achieving sky computing lies in the path to adoption. Stoica and Shenker argue that smaller clouds have the most incentives to embrace this paradigm, potentially compelling larger clouds to follow suit. However, a parallel ecosystem of compute resources has already emerged in the form of layer-1, layer-2, and layer-X blockchains. These decentralized networks of validators provide the compute resources necessary to run application logic, effectively creating “clouds” that integrate compute, storage, and application logic within a single environment.
Interoperability has been a fundamental requirement for these blockchain-based “clouds,” as evidenced by the rise of protocols such as Axelar, LayerZero, and Wormhole. To attract developers and users, blockchains prioritize interoperability from the outset. Additionally, transaction costs are often remarkably low, as the only overhead is incentivizing validators. Interconnectivity between traditional clouds and emerging blockchains could enable hybrid deployment environments, fostering robust competition, affordable execution, and long-term sustainability.
To achieve this, blockchains must support the execution of applications via more traditional software stacks, including databases, DevOps tools, and deployment frameworks. By bridging the gap between traditional clouds and blockchain-based ecosystems, a hybrid environment can emerge to realize the vision of sky computing.
Conclusion.
Computing should be cheap and accessible to everyone. Today’s clouds neither have the incentives nor the tools to realize this vision. A parallel world of blockchains, built from the ground up with global distribution, settlement, and payment primitives, offers an alternative framework for defining the interoperability needed to facilitate sky computing. A permissionless, open interconnectivity standard based on blockchain technology can serve as both an initial launchpad and a long-term enabler for sky computing. In some sense, blockchains are already realizing this vision — developers using blockchains constantly re-launch or migrate their products from one blockchain to another (see the Electric Capital report — “30% of developers support more than one chain, up 10x from 3% in 2015.”). Blockchain technology provides scalable control planes to transform computation into a utility, empowering a competitive ecosystem that ensures fairness, transparency, and accessibility for developers and cloud providers alike.
[1] Ion Stoica and Scott Shenker. From cloud computing to sky computing. In Proceedings of the Workshop on Hot Topics in Operating Systems, HotOS ’21, page 26–32, New York, NY, USA, 2021. Association for Computing Machinery.
Interesting idea of having blockchains support traditional techstacks!