The push for data privacy, combined with the lure of cost savings, has propelled the decentralized cloud to the forefront of technological evolution. Such projects include Definition, OORT, Akash, and more. Unlike a simple spin-off of traditional centralized cloud services like AWS and Google Cloud, the decentralized cloud now lays the foundation for decentralized applications (dApps), the sprawling Web3 ecosystem, and the ever-evolving metaverse. Yet as we venture into this new era, technical challenges, such as auditable computing, become evident.
Verifiable Computing: The Quest for Authenticity
In our world dominated by vast IT needs, outsourcing complex tasks to cloud servers has become commonplace. But therein lies the challenge: once we receive the results, how can we be sure of their accuracy? Consider this: you assign an AI training task to a platform like AWS. A week later, you receive millions of neural network parameters from this AI training task. But how can we ensure that these parameters really reflect the equivalent of a week of training and not just a day?
The simplest solution is to send the identical task to another cloud platform, Google Cloud, and juxtapose the results. However, this method is not only redundant but also doubles the costs. So what is the alternative? This is the topic of auditable IT – a field focused on validating outsourced IT results without re-running the entire process.
Trust problem in a decentralized cloud
In a centralized setup, the problem of verifiable computing still needs to be resolved. Yet consumers rarely question the authenticity of results from platforms like AWS. This trust, whether it comes from brand loyalty or simple lack of alternatives, becomes the de facto solution to the verification problem in centralized clouds.
However, the decentralized cloud universe operates on an entirely different principle: inherent distrust. In this configuration, servers from all over the world can connect. The owners of these servers, driven by their motives, can freely enter or leave this network. This fluidity poses a significant challenge: if you send an AI training task to servers scattered across India, the United States and France without a formal contract or knowledge of their operators, how can you guarantee authenticity results ?
Verifiable IT Solutions
The question of verifiable computing in a decentralized cloud is a challenging problem that researchers have been actively trying to solve for the past 40 years. From these debates emerge innovative solutions based on blockchain, inspired by the Proof-of-Stake mechanism. Essentially, these frameworks require service providers to “stake” or block a certain amount of cryptocurrency when they enter the network. If the network detects foul play or a task issuer raises an alarm, sanctions follow, reducing the assets staked by the malicious party. THE EntrapNet The protocol, which borrows the idea from entrapment in criminal law, serves as a beacon in this context. The key to the successful functioning of such a system is to ensure that it remains decentralized and autonomous. By removing the reins from any entity, Entrapnet fosters an environment where trust can be reestablished in the decentralized domain. In doing so, we move closer to solving the complex puzzle of verifiable computing.
In conclusion, as the decentralized cloud continues to reshape our digital landscapes, it is imperative to address its inherent challenges head-on. Among these challenges, verifiable computing is a long-standing academic problem, which is the key to laying the foundation for a decentralized cloud. Some solutions have been proposed, such as EntrapNet in the OORT data cloud. Yet researchers are studying more effective and simpler responses.