THE STATUS QUO
Data is not useful when it is first generated; it only becomes useful after it has been analyzed. Currently, data is being generated at an incredibly rapid rate – and this rate is only increasing. In order for data to be analyzed, it is then transferred en masse to one of just a few centers around the world. The data that we all generate heads directly to the datacenters of large corporations so that it can be processed – and in turn, monetized.
There are many issues with the current status quo. Two of the most concerning are: (1) we are rapidly approaching the limitations of these centralized data centers’ processing abilities; (2) it is not safe to store all of this sensitive data in the same location – it creates the perfect opportunity for hacking and abuse.
The solution to these issues is to analyze data when and where it is generated. At any given moment, there is so much processing power lying dormant in personal devices such as computers and smartphones. The use of these devices to locally analyze the data that they generate themselves – in a safe manner that prioritizes privacy – eliminates the need for problematic data centers.
So Why Hypernet?
Many projects can connect computers together via smart contracts. Hypernet’s primary innovation is not our on-chain component; it is the off-chain DAC programming model. This model makes it possible to run parallel computations on a dynamic and distributed network of devices, all in an anonymous and privacy-preserving manner. Hypernet brings devices together and uses them to solve real-world problems. Check out our demo video to see for yourself!
What Is Hypernet?
Hypernet is a privacy preserving computing protocol, which can be leveraged across a distributed network of devices.
We have built, from the ground up, a new programming model which we believe will be the backbone of decentralized and privacy preserving parallel computation. Hypernet puts YOU in control of YOUR data, while simultaneously fueling the growing demand for computational power.
Purchase compute power at competitive prices on an infinitely scalable network of devices. Hypernet provides an API suite built on the principle of distributed average consensus and tailored for large-scale distributed compute environments. This API makes harnessing the power of the Hypernet straightforward by encapsulating many practical data science methods (such as clustering, regression, factorization, and filtering) into a single framework.
Earn money by selling unused computing power from the devices you already own. By participating in Hypernet, you are helping solve the world’s greatest challenges by contributing the computational power in your latent internet-connected devices.
Selling time on personal devices dramatically slashes the cost of computing by eliminating cumbersome operating expenses. Upwards of 70% of the operating costs of data centers are infrastructure, hardware, personnel, land, and cooling.
Mobilizing more devices cuts down on compute time. Hypernet is infinitely scalable and can leverage an unlimited number of devices globally. The sheer quantity of latent compute power existing all around us far eclipses that of the largest data center.
Hypernet’s algorithmic efficiency means less energy is wasted by constantly transferring and processing data.
Hypernet makes it possible for massive amounts of data to be computed locally, where it is generated. This eliminates the need to send data back-and-forth to the cloud, saves data-transfer costs, and reduces latency.
The decentralization of data reduces the risk of a major security breach resulting from a single hack. Data decentralization opens the door to a renaissance of data privacy.
Decentralized computing is inherently less vulnerable to disruption or large-scale data theft than a centralized repository.
Hypernet’s DAC API constitutes the next big step in distributed computing technology. It is a parallel programming model that eliminates the need for a “master computer” for compute task coordination. Instead, each computer operates independently. This allows the total system to fluidly add and subtract devices, as needed, to solve a problem. The DAC API introduces a robust solution that can accommodate highly variable distributed environments. Read our whitepaper for more technical details.
Hypernet organizes devices and jobs on the network via the blockchain scheduler. It automatically matches a buyer’s needs with the proper providers, ensures that jobs are completed as efficiently as possible, and helps maintain security and reliability.
The Token Model
Incentives and Governance
Buyers and sellers must stake collateral to complete compute jobs. HyperTokens are that collateral. A seller stakes collateral on their devices while buyers place their payment into the smart contract up front. In a network with unknown actors, the collateral brings peace of mind to both buyers and sellers of compute.
A user’s reputation increases by being a reliable and responsible compute provider and compute purchaser, and this reputation is permanently logged on the blockchain. A user’s reputation increases the likelihood of participating in compute jobs.
HyperTokens are the transactional currency which enables the buying and selling of compute on the network.
Individuals can mine HyperTokens while waiting for compute jobs, by just being available in the lobby. This incentivizes users to join the network and make their devices available. While in the lobby, users can challenge other idle devices to see if they are truly online. If they fail a challenge their collateral is collected by the challenger. The amount of tokens available for mining decreases over time, so signing up devices early earns the most tokens.
Decentralized Governance / Voting
Nodes participate in challenge and response and are incentivized for helping maintain the quality of the network, and weeding out bad actors. Each node pings other nodes in a challenge/response mechanism to determine if they truly are on when they say they are on. Major changes in the network may be voted on, with your vote weighted by the amount of HyperTokens you hold.
Today, the value of data is undeniable. Look at companies like Facebook, their one source of value is their users’ data. Up until now, the extraction of valuable information from all this data was made possible by the extreme centralization of cloud-computing resources. These resources are controlled by a small number of firms, which tends to monopolize data ownership, and monetization of that data. This centralized model proves not only to be technically suboptimal and costly, but also raises fundamental data privacy and control challenges for individuals and society at large.
A new paradigm for data analytics is needed: leave data where it is generated and compute in place. Do not collect user data, and ship it off to be computed in a large data center. Instead, securely bring the code to the data so that it can be computed on device. In such a scenario, the data is stored and computed on the local device to guarantee data privacy, owner control, and security for the end-user, while reducing the data transportation costs. The issue though is that the current infrastructure and programming models available are not well-suited for decentralized computation on device. Therefore, a new programming model is required.
Hypernet introduces a new asynchronous and fault-tolerant programming model to enable general parallel computing with inter-process communication over a dynamic, heterogeneous, and distributed network of edge devices. This is made possible by two distinctly different notions of consensus:
- Consensus in the form of our proprietary programming model: a decentralized computing engine leveraging the mathematical principle of Distributed Average Consensus (DAC) to enable parallel compute tasks. This technical layer combines algorithmic efficiency, robustness under variable network conditions, and privacy-preserving computation.
- A blockchain-based consensus mechanism: to allocate and manage resources by connecting buyers and sellers of compute from the marketplace.
Hypernet’s solution is at the intersection of three related markets: customer analytics & recommendation engines, data protection, and public-cloud, representing 100 b$+ in combined market size, with robust projected growth for the near and long-term future.
Hypernet was founded by a passionate team of three individuals, Ivan Ravlich, Todd Chapman, and Daniel Maren, who met in the early 2010’s during their studies at Stanford University.
- Ivan bridges the divide between various technology disciplines (chemical engineering, materials science, aerospace engineering, theoretical physics) and is a problem solver at heart. He has been involved in many cutting-edge technology startups from bio fuels to plasma rockets and has experienced firsthand the challenges which emerging technologies face. His multidisciplinary experience gives him foundational knowledge to develop an intricate a multidisciplinary product like Hypernet.
- Todd lives and breathes parallel computing methods. Throughout his PhD research at Stanford, Todd was able to discover deep truths of applied mathematics that have allowed him to develop Hypernet’s unique decentralized computing methodology. He leads the technical team with his vision of the overall architecture of Hypernet.
- Daniel is a serial entrepreneur who got his start studying CS at Stanford. He left early in his studies to form a solar company that was later acquired by Sun Power. He brings both the operations and business development skills from his experience in startups and large corporations.
The founders and the team they have assembled will scale Hypernet through all the stages of its growth to build an organization that can change the face of the computation and data privacy through its technology and culture.
The team is currently composed of 9 full-time members, half of them with robust business experiences, including: (1) Daniel Maren, who co-founded a company acquired by SunPower Corporation, (2) Christopher Hansen, who was a software engineer at Facebook and Nvidia, (3) Samy Biyadi, who invested in and advised several tech startups, in the US and in Europe, especially in the fields of Smart-Energy, SaaS, and Smart-Mobility. The other half of the team has been in PhD programs, with intensive use of numerical methods, or in PR/marketing.
The extreme centralization of cloud-computing resources, which includes both large amounts of computational power and personal and social datasets in the hands of a few firms, raises data privacy, governance, and control challenges not only for individuals but also for society as a whole. From a data security perspective, the more data is centralized, the higher the incidence of malicious attacks as the reward from a single successful hack is more valuable. As such, data breaches have reportedly shown that millions, and sometimes billions, of customers were affected by a single hack, most often negatively impacting the financial valuation of the associated companies.
The Hypertoken is needed for four interconnected and critical functions to ensure a healthy and transparent marketplace:
- Collateral: buyers and sellers of computational power will use tokens as a form of collateral for launching and fulfilling compute jobs. Using tokens as collateral will ensure fairness for all parties and limit bad actors from misusing the marketplace.
- Network Allocation: Hypernet tokens will be used to allocate and measure capacity for submitting and running compute jobs.
- Reputation: Hypernet tokens will allow buyers and sellers to build a verifiable reputation, and increase their standing to be able to request and fulfill compute jobs.
- Priority: The prioritization of jobs will be moderated within the Hypernet by the previous three factors of reputation, network allocation, and collateral.
In the marketplace, we intend to work with 3rd-party providers to enable FIAT payments, and crypto-currency payments with major crypto-assets (NEO, EOS, ETH, BTC, XMR…). This is inline with Hypernet’s goal of providing a simple and elegant user experience in order to help with adoption.
A buyer should be able to post and execute a job without interacting directly with HyperToken (it will all be handled on the backend away from the buyer). This is a key requirement to on-board corporations with regulatory requirements that do not allow them to buy and own tokens.
We have designed the concept of a lobby (a ‘holding area’ for available CPUs) in order to ease the management of supply and demand. While in the lobby, CPUs are rewarded with tokens just for being available (up-time rewards), even when not executing a computing job. The reward will be proportional to their relative computational power contribution to the total computational power of the lobby.
To enter the lobby, a supplier A stakes collateral. This collateral could be lost if another supplier, B, detects that supplier A is not actually available. Availability is determined by a simple computational challenge that should be statistically solved in a certain amount of time. If supplier A fails this test, then its collateral would be transferred to supplier B.
This mechanism incentivizes user to join the network, while disincentivizing bad/malicious actors.