Decentralized Delivery of Videos: GPUcoin White Paper

ABSTRACT

GPUcoin is building the first comprehensive protocol for the decentralized delivery of videos. At the core of its patent-pending approach, GPUcoin removes a number of high-cost processes to deliver video in real-time. Today, video delivery requires integrating expensive video players, streaming servers and encoders from third party companies (such as Adobe) and delivering the content over expensive, centralized delivery networks “CDNs” (such as Akamai) to speed up live video playback. The existing process not only depends on a decades-old architecture with expensive licensing costs, but it suffers from high latency in its delivery. GPUcoin’s completely new approach eliminates the need (and high costs) for a third party CDN, video player and streaming server. Instead, it will depend on a new architecture and use a decentralized peer-to-peer network to provide near real time delivery. Utilizing advances in blockchain technology along with a distributed network of participants’ Graphic Processing Units (GPUs), GPUcoin will eliminate these expensive content delivery networks for streaming video in news, education, sports, gaming and music. This live-streaming protocol will be incentivized by GPUCoin mining authenticated by Proof of Streaming (PoS), using non-interactive zero-knowledge proofs (zk-snarks). For the first time, users will be able to stream video as quickly as their internet connection will allow.

A NEW INTERNET AGE

GPUCOIN’S PLACE IN THE THIRD EVOLUTION OF THE WEB

Over the past few decades, we have had the privilege of watching the Internet grow from two nodes communicating between UCLA and Stanford into a vitally-important global network of data centers, companies and users.

As the internet grew from its infancy in university research labs and military installations, more and more nodes were added to the network, eventually powering a decentralized global community. As technology evolved, a web stack composed of building blocks like TCP/IP, DNS, databases, web servers, authentication systems and CDNs were developed. Many great businesses were then built on these new network protocols, over time growing so large and omnipresent that a select few have managed to take central control over many of the internet’s primary functions. Today, the vast majority of the financial transactions, communications, data collection, advertising and hosting that occurs on the internet goes through an Amazon data center, a Google app or an Apple iPhone. This concentrated centralization is a key hallmark of the second age of the internet, by-products of which include significant risk to our privacy and security as well as out being held hostage to a third party’s unilateral pricing power. This is a major reason why there is such an appetite for the web to evolve again.

This new and third evolution of the internet has been referred to as Web 3.0 or Web3. Those advocating for the internet’s return to a decentralized model range from open-minded idealists to entrepreneurial technologists and investors. While their reasoning may differ, they agree on the fact that blockchain technologies are offering a fairer model for businesses and users that want to interact on a common network without relying on a central authority to handle a transaction, manage and store their data or share a piece of content.

With the advent of the Web3 movement, we have entered a new era of decentralized applications made possible by blockchain technologies and cryptocurrencies. In the centralized web, the users are the product – their interests, preferences and profiles are managed by companies such as Facebook, Twitter and Google and then packaged and sold to advertisers, enriching a small group of shareholders, who help create significant barriers to entry in the marketplace. In the decentralized Web3, user information is private, and value creation is not primarily about advertisements and shareholder enrichment.

Solving this problem will require a new set of “foundational” protocols to become the building blocks of this Web3. Some new protocols are already emerging. Bitcoin seems to be on its way to become a store of value. Ethereum has a head start as a smart contract protocol. Add to that the thousands of other foundational blockchain protocols are being considered and or built, and you begin to see the tremendous potential of these technologies.

It is with this understanding that we introduce GPUcoin, a global network protocol for streaming video-enabled DApps. The GPUcoin platform will provide a cryptographically secure, fault-tolerant, self-amending distributed computing environment. GPUcoin’s Real-Time Satoshi Streaming Protocol (RTSSP) will use peer-to-peer nodes for delivering both live-streaming video, and archived videos, verified by Proof of Delivery (PoD). RTSSP is a peer-to-peer hyper stream protocol that makes live-streaming & distributed computing on GPUs more secure, faster and safer, as well as more open and censorship-resistant.

This streaming video protocol will be able to deliver content with both lower latency and cost compared to the current centralized model, allowing users to stream video as quickly as their Wi-Fi connection will allow.

VIDEO STREAMING MARKET OPPORTUNITY

THE CURRENT DIGITAL MEDIA MARKET GENERATED GLOBAL REVENUE OF MORE THAN $100 BILLION IN 2017[1]

Approximately a fifth of that was video – in particular, streaming video. And more growth is on the horizon. However, these figures do not account for the impending growth in augmented and virtual reality demand. To assess the potential impact of VR and AR formats might have on the market, we reviewed several recent trends and forecasts on video storage demand:

According to the 2016 Digital Storage in Media and Entertainment Report from Coughlin Associates, “the rise of VR, and IoT technologies in and around the home, data movement within and outside the home will increase. It was estimated that the average internet user will be responsible for about 1.5 GB of internet traffic per day and generate about 7.5 GB of data per day by 2020.”[2]

When writing about the potential impact VR might have on data storage, Data Center Frontier reporter Rich Miller noted that, “these 360-degree video [VR] applications require a LOT of data, and delivering these experiences across the Internet presents a major challenge. A video at the 6K resolution used by the GearVR headset could be 20 times the size of today’s full HD video. All that data will have to move across the network, and in some cases be cached locally to assure low latency.”

In the same article, Jay Parikh, Global Head of Engineering and Infrastructure at Facebook noted that “by 2020, more than a zettabyte – that’s 1,000 exabytes – of information will be exchanged over telecom networks, much of it in data-intensive formats like video and virtual reality.” He continued, “Scaling traditional telecom infrastructure to meet this global data challenge is not moving as fast as people need it to … We know there isn’t a single solution for this, and no one company can tackle the problem alone.”[3]

Compared to “structured” file storage infrastructure (like spreadsheets and databases), which is not optimal for video,[4] storage demand for “unstructured” data (like videos, images and audio) is set to experience significant additional growth in the coming years.[5]

To support this massive growth, Web3 requires fat protocol level development to enable distributed streaming video, including VR and AR, at scale. Foundational protocol layers like FileCoin & GPUcoin will form the next Web 3.0:

Phase	Internet	Cryptocurrency	Reach
Protocol	TCP/IP, SMTP	e-gold, digicash, bitgold, Bitcoin, Ethereum	1M People
Infrastructure	ISPs, lay fiber	Exchanges, secure storage	10M people
Consumer Interface	Browser	User controlled BTC/ETH wallets	100M people
Decentralized Ðapps	Web 2.0	Finance 2.0 Ðapps	1B people
Fat Protocols	Web 3.0	FileCoin, GPUCoin empower next-gen Ðapps	2B people

PROBLEMS WITH THE CURRENT CENTRALIZED MODEL

There are both centralization and technical problems associated with the current live-streaming system, including centralized points of failure in backend streaming software and content delivery networks. These technologies are also proprietary or closed source software which means they are prone to censorship and easy to control or shutdown the service. Consider expensive licensing fees and high latency and you find a system in desperate need of innovation.

WE PLAN TO ADDRESS FOUR KEY WEB 2.0-ERA PROBLEMS WITH GPUCOIN:

1. Existing entertainment systems are designed to benefit a select few companies and their leaders. Most artists, musicians and creators have little or no access to the monetary benefits of their own creations.
2. Internet Service Providers (ISPs) currently have significant control, including the option to slow down speeds or block content, applications or sites you may wish to use.
3. Most video and live streaming systems are centralized and subject to censorship, leading an unwillingness or inability for citizens to share their opinions and voice.
4. Due to their inherent centralization, content delivery systems are expensive and lack innovation. This is in part because existing monetization systems are rudimentary with intrusive advertisements that are hard to avoid when consuming streaming content.

EXISTING SYSTEM ARCHITECTURE:

The earlier illustration shows the current live-streaming system. There are four components to a live-streaming system. They are explained below, for context as we begin to present our GPUcoin solution.

PART 1: BROADCASTING

A proprietary mobile video encoding software on an Apple iPhone, for example, is used. The primary purpose of this software is to capture video frames and audio, from mobile, desktop, or stand-alone cameras. The software encodes the captured video and audio frames into a video standard, and a closed/proprietary video streaming protocol and is published to a centralized streaming software platform.

BROADCASTING SERVER SOFTWARE

Current software that solves this problem are Wowza[6] and Adobe[7]. The Broadcasting software receives the encoded live streams and generates small fragments of video files that are then published to a Content Delivery Network. This software is expensive and has significant latency.

PART 2: BACKEND CONTENT DELIVERY NETWORK

The Broadcasting Server Software publishes the generated video fragment files to Content Delivery Network such as Amazon Cloud-front[8] or Microsoft Azure.[9] These backend networks are also expensive to use and add additional latency to the streaming video process.

PART 3: PLAYBACK / VIDEO PLAYER

This is the final step in live video streaming. Media/Video/Audio clients for platforms: mobile, desktop, play out the video files from the content delivery networks.

THE GPUCOIN PROCOTOL

AN OPEN, DISTRIBUTED MARKETPLACE IS THE SOLUTION TO MANY OF THE PROBLEMS THE OLD SYSTEM FAILS TO ADDRESS.

Streaming video services still depend on centralized computing resources for content delivery. The GPUcoin Network will act as a marketplace, allowing anyone to join the network both as a Provider – selling unused GPU capacity, or as a Customer – buying live-streaming services from GPUcoin Network providers. Adoption and use of the network will be incentivized by a crypto-economic model that supports network actors and encourages innovation. GPUcoin is committed to:

1. Giving creators a way to monetize their own creations. This will empower them to create, trusting the open decentralized marketplace to fairly compensate them over existing centralized channels.
2. Supporting strong net neutrality by design. Meaning the GPUCoin decentralized network prevents any form of throttling, blocking or slowing down of content.
3. Making the video system decentralized and censorship resistant. The GPUcoin network will promote free speech without risk of detection and censorship.
4. Assembling video systems that are more expressive. Because these video systems will be democratic and censorship resistant, it will hopefully lead to a new vein of video innovation on the open decentralized web.
5. Making the platform open source and decentralized over the blockchain. This means creators can price their content and accept cryptocurrencies in place of intrusive advertisements which primarily benefit the current centralized oligarchy.
6. Developing the best platform in the decentralized world for live streaming video. The GPUcoin protocol will help launch new, innovative ideas through streaming-enabled Distributed Applications (DApps)

PARTICIPANTS IN THE GPUCOIN NETWORK

GPUcoin Video Streaming Process Example

The following process loosely defines how a video will travel through the network:

1. Streamer stakes GPUcoin to claim use of the network (on-chain)
2. Streamer uploads video through GPUCoin’s open source Broadcasting software
3. Viewer requests to view video
4. Miners are selected from those available based, in part, on the size of their delegated stake to stream the content
5. Streamer is notified that the request has been accepted, at which point the video transfer between streamer and miner occurs
6. Miner delivers video to viewer through GPUcoin’s open source video player in real time. Video is also archived in IPFS.
7. Proof of Streaming verified using zk-SNARKs (off-chain)
8. Once Proof of Streaming is verified, the wining miner writes the result of the stream on-chain
   1. If the miner fails verification, funds are slashed from that miners staked funds and returned to the streamer
   2. If verification is passed, the miner who finds the winning block earns the GPUCoin reward (all the other competing GPUCoin streaming nodes earn the transaction fees)

STREAMER NODES

[In order to participate, streamer nodes must stake GPUcoins]

Streamers may:

1. Request streaming capacity on the network through a staking transaction
2. Send video to miners through GPUcoin’s open source live video encoder
3. Receive micropayments from viewers for paid content

MINER NODES

[In order to participate, miner nodes must stake GPUcoins]

Miners may:

1. Offer excess GPU capacity to network participants through staking transactions
2. Request/receive streaming video work based on proportion of staked GPU coin
3. Earn GPUcoin transaction fees as reward for verified, complete work

VIEWER NODES

[In order to participate, miner nodes must stake GPUcoins]

Miners may:

1. Request video from the network
2. Open pay-per-view payment channels with streamers for premium content
3. Receive video from the network’s miner node

INTERACTIONS IN THE GPUCOIN NETWORK

The diagram above is the proposed GPUcoin system. This model is further explained below. In its simplest form, the GPUcoin protocol facilitates an ecosystem in which anyone with excess GPU capacity can receive payment in cryptocurrency from users who come to the network to watch the video they deliver. The GPUcoin protocol integrates all three pieces of traditional video delivery into a crypto-economic system that distributes both the work (video delivery) and reward (content or payment for content) more evenly amongst its participants.

GPUCOIN SUPPORT SYSYEMS/PROTOCALS

GPUCOIN CRYPTOCURRENCY

The GPUCoin network will incorporate a native cryptocurrency called GPUCoin (GPC). GPUcoin will be used to pay miners for the use of their excess GPU power. Miners running nodes on the network will stake GPUcoin through a delegated proof of stake (DPoS) consensus model, which we will discuss in greater detail later.

The GPUCoin Network will apply the Work Token Model pioneered by projects like Augur and discussed extensively by thought-leaders like Multicoin Capital. The Work Token model will, in part, ensure that price of the token increases linearly with the usage of the network.[10] By strengthening some of the incentives to hold GPUcoin, rather than simply offloading it once it has done its work, we will be able to capture more value for the networks’ participants than a utility token model.

GPUCOIN SOFTWARE INTEGRATIONS (HOOT)

IPFS FOR ARCHIVED VIDEOS

Broadcasted video will be continuously archived and stored in the decentralized file system IPFS.[11] This means that the GPUCoin live stream protocol and file archival system makes live-streaming faster, safer, and more open through:

• Having no single, centralized points of failure in highly resilient network with Peer-to-Peer nodes (Byzantine fault tolerance).
• Censorship resistant as there are no centralized points of control.
• The open source nature of its core networking software.

PROOF OF STREAMING WITH ZK-SNARKS

As GPUcoin progresses, we will be exploring options for verification of work, focusing first on using non-interactive zero-knowledge proofs for Proof of Streaming. Several other options may be explored, including but not limited to those described in the Livepeer whitepaper (Truebit, 3^rd party computations services and hardware enclaves).[12]

DELEGATED PROOF OF STAKE CONSENSUS AND SLASHING

In delegated proof of stake, anyone who holds GPUcoin is able to “vote” in support of trusted block producers, incentivizing higher quality service from miners. The greater the relative delegation supporting a miner, the more GPUcoin that miner is able to produce. While this consensus model sacrifices some decentralization, it allows for better scalability and a transparent governance model to develop.

Additionally, because slashing functions can easily be implemented under DPoS (in which some of the staked GPUcoin is taken or “slashed” from a miner who hasn’t delivered on the terms of the GPUcoin smart contract), there is a genuine threat of loss of income for poor performance. This incentivizes miners to provide the bandwidth and GPU power required to meet viewer and streamer demands, along with creating competition between miners for rewards.

Example Probabilistic Payment Scenario developed by Orchid Co-Founder and former Ethereum Core Dev Gustav Simonsson:

• You are the video streaming host, charging $1 per hour of video streamed.
• A new user connects to your service and watches exactly 10 seconds of cute kittens.
• Every 4s, the user sends you an off-chain ticket with 1/2500 probability of winning. The winning amount is set to $2.78.
• If you do not receive a payment ticket every 4s (within some margin to account for network latency), you simply disconnect the viewer
• If the viewer does not receive any video feed, they immediately cease sending payment tickets.
• When you receive a winning ticket, you claim it with an on-chain transaction, which transfers the wining value to you from the tokens locked up by the sender.

Even if a ticket is not winning, you still have cryptographic proof that you’re being paid, since neither the sender nor the recipient can manipulate the value that determines if a ticket wins. With the proof, you’re happy to continue streaming video to the viewer. As thousands of (new) viewers connect to you, there is no overhead per viewer —  the only overhead is when you claim your winning tickets. For a given “average total payment”, the odds of winning (and frequency of tickets) determine the number of on-chain transactions required. Thus, the higher the winning amount, the less transaction fees paid over time by the recipient. The lower the winning amount, the less tokens have to be locked up by the sender, lowering their liquidity cost.

PROBABILISTIC MICROPAYMENTS

GPUcoin network will leverage probabilistic micropayments to overcome channel initialization transaction fee issues associated with the settling small payments on the Ethereum blockchain. This also helps alleviate some of the current scaling issues associated with Ethereum.

Ethereum Probabilistic Micropayments will allow GPUcoin to send small payments between viewers and streamers without a huge number of per-recipient initialization or settlement transactions. These micropayment channels will allow the GPUcoin network to execute a number of transactions with users while only requiring two on-chain transactions: one transaction to initialize the payment channel and one transaction to close the channel and settle total amounts transferred.[13]

GPUCOIN MATH AND TOKEN ECONOMICS

The following section describes the math behind GPUcoin as well as how it will compete economically against other streaming solutions as well as other minable coins in the crypto ecosystem.

GPUcoin Math

Running the lowest cost unit containing a GPU, an AWS p2.xlarge, costs about awsdc = $0.9/hr. This unit is equipped with a Tesla K80 GPU or similar GPU, along with 32GB or 64GB RAM and a 4 core CPU. Hence assuming the dollar cost of an average AWS[14] GPU instance per hour at about 1$ per hour is a fair assumption. [15]

The monthly dollar cost of AWS GPU instance is p = awsdc*24*30 = 1*24*30 = 720$ [1$ per hour per aws GPU instance * 24 hours per day * 30 days per month]

The GPUCoin instance cost is pegged at 1/5th AWS costs to stay competitive and as GPUCoin nodes do not pay infrastructure and electricity costs in the decentralized model. This can be generalized as peg variable that GPUCoin miners can tweak based on the availability and scarcity of GPUs and render farms in their region. Thus, we recommend pegging to fiat only in the short term as a bootstrapping mechanism. In the long term, this variable will be dynamically adjusted using a vickrey auction, eliminating the dependence on fiat currency, while maintaining decentralization, efficiency and economic price stability.
peg = 0.2

The dollar cost for a GPUCoin instance comparable to an AWS GPU instance

gpcdc = peg*awsdc = 1/5* AWS cost = 1/5 * 1$ = 0.2$.

Hence 0.2$ is the dollar cost per comparable GPUCoin instance per hour as the GPUCoin IPCN network does not have to pay for infrastructure and real estate, these are sunk costs for the GPUCoin network miners. This model holds true even if there is no peg and we assume about 1/5th idle network utilization, i.e., the network utilizes the GPU null cycles for 1/5th of the time.

The monthly dollar cost of a comparable GPUCoin instance is

q = gpcdc*24*30 = 0.2*24*30 = 144$

[ 0.2$ per hour per GPU instance * 24 hours per day * 30 days per month ].

Taking t = 20$ for transaction fees and margins

GPUCoin mining revenue per month for 1 GPUCoin instance r = q – t = 144-20 = 124$

Ethereum mining revenue from one AWS GPU instance
e = 31$

r = 4* e

Mining Profitability

Based on these calculations, GPUCoin mining is four times more profitable than Ethereum mining (which is a profitable GPU mined cryptocurrency), even at 1/5^th network utilization. This is a reasonable profit for something that takes no effort, and comparable to Bitcoin mining profits during the golden GPU mining era circa 2009-2011.

Initial GPUCoin token difficulty for mining follows the bitcoin mining model i.e., the token max limit is a hard cap of c = 21 million GPUCoins to honor Satoshi Nakamoto’s Bitcoin. Early decentralized GPUCoin nodes earn more GPUCoins as a block reward as there are fewer nodes in the network; Proof of Streaming difficulty increases with the number of GPUCoin nodes in the network.

Whereas earlier 1,000 GPUCoins would be needed to buy time in 1 single GPUCoin instance, later as the network size increases as more nodes join the GPUCoin network 1 single GPUCoin would be able to buy time in 1,000+ GPUCoin nodes few years down the line as GPUCoin tokens appreciate in value due to scarcity of mining, a hard cap of c = 21 million GPUCoins and strong sustainable demand for GPUCoin instances that will be generated by the cheaper than AWS pricing of a comparable GPU instance.

An example scenario is when there are just 10 GPUCoin nodes in the network the Proof of Streaming difficulty is roughly 1 hour of streaming needed to earn 1 GPUCoin as block reward. As more GPUCoin nodes join the GPUCoin network the PoS difficulty goes up, when there are 10,000 GPUCoin nodes in the network the mining difficulty is about 1000 hours of streaming needed to earn 1 GPUCoin as block reward.

Irrespective of the GPUCoin price and mining difficulty the GPUCoin instance dollars costs are recommended to be cheaper than their corresponding AWS GPU instance dollar costs. This ensures that the GPUCoin dollar costs stay competitive and attractive compared to the AWS instance dollar costs in order to drive sustained GPUCoin instance demand.

As mining difficulty is adjusted based on the number of the nodes compute capacity that the network controls and the maximum number of GPUCoins that can ever be mined is hard-capped at 21 million GPUCoins the GPUCoin cryptocurrency is designed to be a deflationary currency relative to fiat currencies similar to bitcoin and litecoin. While the miner who finds the winning block earns the GPUCoin block reward all the competing GPUCoin streaming nodes earn the transaction fees. Hence in Bitcoin there is one winner and multiple losers whereas in GPUCoin there is one winner and no losers i.e., multiple miners sharing mining rewards corresponding to their Proof of Stake or the number of GPUCoin nodes they used for PoC or PoS mining.

INSURING INFLATION RATE DOES NOT OUT-PACE GROWTH OF THE UNDERLYING ECONOMY

The sum total of all the GPUCoins GPC coins minted at each interval can be guaranteed to be less than the calculated rate of economic growth similar to the variable bitcoin hash-cash difficulty innovation, such that current token holders can be better assured that prices are not likely to inflate and that their tokens will fall below their original value because of an over inflation of excess GPUCoin Token supply.

A further percentage of the growth can be retained as explained above to provide a price floor as an assurance of value maintenance to GPUCoin Token holders.

ERC223 COMPATIBILITY

We are monitoring the ERC223 token standard proposal[16] and are factoring future compatibility into the design of our Namespace GPUCoin tokens.

CHOICE OF BLOCKCHAIN

The seed protocol will run on top of the Ethereum blockchain protocol which makes the decentralized app abstraction. We take advantage of Ethereum blockchain technology to ensure a fair democratic usage of the network for all participants miners, consumers and users of GPC GPUCoin tokens.

GPUCOIN CARBON FOOTPRINT, MINING, SCARCITY AND PROFITABILITY

Because excess bandwidth and compute capacity will be put towards streaming, encoding, object recognition and security of video and audio streams, the resources that would otherwise be underutilized are instead made profitable.

Since GPUCoin tokens are fairly distributed/mined to miners corresponding to their compute and bandwidth availability, the GPUCoin carbon footprint will be much lower than the Bitcoin network, which depends on a continuously increasing hashing complexity for mining. Since GPUCoin tokens can only be mined or acquired from the platform and since they have a hard cap of 21 million GPUcoins, they will tend to be a scarce and valuable deflationary cryptocurrency token.

MERKLEDB: GPUCOIN’S CENSORSHIP RESISTANT PEER-TO-PEER IDENTITY, TRUST & REPUTATION BLOCKCHAIN

Since there is an economy of trading in the marketplace of the GPUCoin network, having a Peer-to-Peer identity and reputation database MerkleDB to enable seamless, non-custodial decentralized, trust-free interactions becomes essential. Users or Agents may be identified using platforms like Civic, UPort, or using the ecosystem the Decentralized Identity Foundation[17] is building.

Feedback and reviews, as well as scoring on a 1-5 scale for quality of interactions factor into an agent’s reputation trust score. The trust score of each agent is hashed into the blockchain using their public GPG key and hashed user-name or decentralized identity so as to make them censorship resistant.

Trust scores and reviews may only be added to the database – nothing can ever be removed – making this the trust reputation blockchain. Merkle trees, an efficient verifiable data-structure, are used to ensure the reputation database is usable by making it possible to download the relevant sub-tree for a particular user’s sub-network hash even as the reputation blockchain grows.

MULTI-SIG ESCROW WALLETS

GPUCoin tokens are first sent to a multi-signature escrow wallet that is controlled by the buyer (viewer), seller (streamer) and an independent 3

rd-party escrow. Any two out of the three parties need to sign with their private keys for the transaction to be completed. Also, the number of times the buyer or seller necessitates escrow agents to mediate a dispute and the time to complete a transaction will factor into the reputation of the buyer and seller.

Any trusted agent with a high enough reputation score can register to be an independent 3

rdparty escrow agent. Escrow agents also earn feedback and trust which are hashed and stored in the block-chain using their public GPG key and hashed username, so it becomes censorship resistant.

VICKREY AUCTION TO FIND OPTIMAL PRICE

To bootstrap the network, GPUCoin IPCN will run Vickrey auctions to find the best service to run on miners’ excess GPU capacity.

A Vickrey auction is one in which the winner pays the second-highest price, not the price they themselves bid, which has been effectively used by Google Adsense and Adwords. GPUCoin can implement any auction protocol, if we find a suitable auction protocol that is superior to Vickrey.

Upgrades to GPUCoin auction protocol are staged through a testing environment to allow stake-holders and token-holders to recall potentially inferior amendments, that lead to sub-optimal pricing for network stakeholders.

Since the GPUCoin IPCN network may also be used for other tasks than streaming live video, the network can be extended to run any distributed computing task such as computer graphics, business applications, machine learning, cryptography, malware prevention analysis, science and services, making the GPUCoin IPCN network a Uber for computers, enabling miners to rent their unused CPU/GPU cycles & get paid in GPUCoin crypto-currency. Hence the GPUCoin decentralized network powers true distributed cloud computing accelerated by GPUs.

HOOT: LIVE VIDEO DISTRIBUTION

ULTRA LOW LATENCY VIDEO PROTOCOL

Hoot Live is building the first protocol for decentralized delivery of videos. At the core of its patent-pending approach, Hoot Live removes a number of high cost processes to deliver video in real-time.

Today, video delivery requires integrating expensive video players, streaming servers and encoders from third party companies (such as Adobe) and delivering the content over expensive, centralized delivery networks (“CDNs”) (such as Akamai) to speedup live video playback. The existing process not only depends on a decades old architecture with expensive licensing costs, but it suffers from high latency in its delivery.

Hoot’s completely new approach eliminates the need for and high costs of a third party CDN, video player and streaming server. Instead, it will depend on a new architecture and use a decentralized peer-to-peer network to provide near real-time delivery. Through Hoot, users are finally able to stream video as quickly as their internet connect will allow.

TECHNOLOGY BENEFITS

When compared to other products on the market, this has several defensible advantages:

• Optimized for 2G/3G networks around the world, low CPU and GPU usage, saves battery and bandwidth consumption
• Hoot’s next-gen live-streaming product enables mobile phone self-serve streaming
• Allows interleaving of background music seamlessly
• Breakthrough patent-pending technology designed from the ground up, requiring no licensing fees
• Hoot’s livestreaming app offers instant archival of live-stream videos making it unnecessary to upload files again at the end of live-stream
• Modular architecture allows building Tor/VPN modules in order to enable censorship resistant live-streaming to promote free speech

HOOT TRACTION & USAGE

Hoot is a mobile consumer application is currently live on the iOS AppStore and Google Play-store. Hoot will provide the basis for the GPUCoin Open Source Video player. Hoot has received streams from:

tables [table:1] and [table:2] show the usage statistics. distribution User Statistics

Monthly Active Users (MAU)

214,769

Live Video Streaming Statistics

Number of Videos	48,207
Average Viewers per stream	155
Average Stream Duration	4 minutes 35 seconds
Total Time Watched	37k days or 100 years

GPUCOIN VIDEO PLATFORM, PLUGINS AND VIDEO APPSTORE

We are quite excited by ARKit and ARCore which are going to be soon available in the upcoming versions of iOS11 and Android respectively.

We believe this framework is going to support great advancements in augmented reality video-streaming. With Tensor-flow, Caffe, Keras, Theano, TensorFlow Object Detection API, GCP cloud video intelligence API and Azure video machine-learning APIs, there is an impending wave of machine learning video content-analysis applications, making videos fully searchable and discoverable. You will soon be able to search every moment of every video file in your catalog and find every occurrence and its significance. IPCN GPC video primitives via API/SDK will allow developers to extract actionable insights from video files without requiring any machine learning or computer vision knowledge.

We envision an Augmented Reality-first OS, an operating system native to virtual and augmented reality. We want to enable a thriving ecosystem by building a video App Store and provide a scalable platform for video developers around the world. By enabling a platform with easily extensible and scriptable plugins, and video App Store for AR and ML/AI v-apps, we will accelerate this new age of intelligent live video and augmented reality. By providing an extensible plugin architecture, we will allow developers to build plugins on the IPCN video architecture. Filters, face-detection and face-swapping are some early augmented reality ideas that will be explored.

Some other interesting machine-learning and AI ideas that would likely be explored include Label Detection (detect entities within the video, such as “dog”, “flower” or “car”), Shot Change Detection (detect scene changes within the video), Regionalization (automagically specify a region where processing will take place), automated subtitle detection, home/office security & other applications yet to be discovered may also be encouraged on the GPUCoin distributed computing platform.

The underlying GPUCoin technology may also be used to build an open source low cost security & surveillance alternative to closed systems such as Nest.

RTSSP ARCHITECTURE

BROADCAST SIDE – MOBILE IOS CLIENT

The protocol for real-time livestreaming video is called Real-Time Satoshi Streaming Protocol [RTSSP]. Video frames are captured at a resolution of 540×960 to 720×1280 based on network connectivity. Audio streams are captured using the built in iOS device microphone at a sampling rate of 44.1 KHz.

Optionally, real time filters (Black and White, Glow, Fisheye, Sepia) can be applied to captured video frames in real-time.

Video and audio are encoded using the native hardware H.264(H.265 in android) and AAC encoders, respectively. The video frames are encoded using a VBR algorithm with a maximum bitrate of 1 Mbps, this can be increased for use cases such as VR streaming. Audio stream is encoded in AAC format with a bitrate of 128 Kbps. The H.264 + AAC stream is encoded into an RTSSP stream and is transmitted to Hoot RTSSP server.

BROADCAST SIDE — DESKTOP MAC CLIENT

Video frames are captured at native screen resolution, and audio stream is captured using the built-in microphone at a sampling rate of 44.1 KHz. The Hoot native cocoa Mac app written in Objective-C supports capturing FaceTime, Screenshare, and a combination of FaceTime and Screenshare.

Video frames and audio stream are encoded using the native H.264 and AAC encoders, respectively. The video frames are encoded with a VBR algorithm. Audio stream is encoded in AAC format with a bitrate of 128 Kbps. The H.264 + AAC stream is encoded into an RTSSP stream and is transmitted to the open source RTSSP server.

VIEWER SIDE MOBILE IOS/ANDROID CLIENT

Hoot open source Native mobile media player decodes RTSSP + H.264 and AAC data to make the live broadcast available to viewer in real-time. The HLS (HTTP Live Streaming) stream that is made available can be played using the iOS/Android Native media players, when the Hoot RTSSP player or app is not available.

VIEWER SIDE MAC/ DESTKOP PC CLIENT

The RTSSP stream is played using Adobe Flash technology supported by modern browsers. The HLS stream can be played using HTML5 player available in modern browsers.

SERVER-SIDE PEER-TO-PEER DECENTRALIZED TECHNOLOGY

Similar to Bitcoin blockchain technology, any node can join or leave the Hoot network at any time. Each node runs a real-time broadcasting server.

The GPUCoin network has several RTSSP servers that serve to bootstrap the Network. We use commodity servers with modern processors and with 1 Gbps duplex Ethernet; specialized servers are not needed.

The hoot server generates two variants of streams: a RTSSP stream and a HLS stream in order to make them accessible in browsers across Windows, Mac OS, Linux and Android platforms. A server with 1 Gbps duplex ethernet can support up to a total of 1,000 viewers. A stream is replicated horizontally across multiple servers (without additional latency) to stream to virtually an unlimited number of simultaneous viewers. Streamed videos are instantly archived [H.264+AAC, mp4 container] in the cloud for later viewing. The archived videos are indexed (scrubbable and quick to scan).

We have access to datacenters in the following geographically distributed locations through RTSSP servers to provide the least latency to viewers globally: Amsterdam Netherlands, Frankfurt Germany, Hong Kong, London UK, Melbourne Australia, Queretaro Mexico, Milan Italy, Montreal Canada, Toronto Canada, Paris France, Singapore, Sydney Australia, Tokyo Japan, Dallas TX, Houston TX, San Jose CA, Seattle WA, Washington DC. Streams are replicated and pulled to the closest node to the viewers location, i.e., a viewer in Tokyo Japan viewing a stream from Washington DC would be connected to a replicated stream on the Tokyo Japan hoot node in order to reduce latency.

SECURITY

HOOT

The live connection is encrypted using AES_256_CBC, with HMAC-SHA1 for message authentication and DHE_RSA as the key exchange mechanism. Every Hoot open source player connection is authenticated. An authorization key is needed to view a private Hoot video stream. Signup, interactions, HLS streams and archived static content are end-to-end HTTPS SSL encrypted to ensure strong security.

ANONYMITY AND PRIVACY OVER VPN AND TOR

Anonymity and privacy are key to enable free speech, and this matters more in countries where free speech continues to be an ongoing human rights issue. In combination with blockchain technology, the network is designed to route video streams and meta data over VPN and optionally Tor network to avoid censorship and promote free speech.

SYBIL ATTACKS

Requirements to stake GPUcoin in order to gain influence in the network provides a strong economic barrier to this kind of network attack. However, the GPUcoin team will be exploring additional protections from Sybil attacks through Proof-of Streaming using zk-SNARKs.

ENTERPRISE AND CONSUMER BENEFITS

ENTERPRISE CUSTOMER PROBLEM

Large, distributed teams in complex and highly regulated environments naturally create pressure on their meta organizations to support better communication, collaboration, and knowledge capture as they evolve.

With rapid innovation in online and mobile video formats and in video delivery technologies, enterprises that wait years between major platform updates will frustrate customers and employees with outdated experiences. The current live-streaming products are too complex and expensive, leaving customers with cobbled-together solutions that don’t work well and require continuous IT support.

Current providers are neither mobile-first nor self-serve, while also suffering from significant latency, high bandwidth, and battery consumption issues. As major browsers no longer support plugins, user experience is sacrificed.

Example enterprise customers are Zoom, Livestream, Twitch, Salesforce, Coursera, Udacity and Stanford University, among others.

ENTERPRISE SOLUTION / PRODUCT OFFERING

GPUcoin will dramatically reduce the pain of running an enterprise scale live-video stream by offering a next generation, powerful platform that:

• Streamlines the communication and collaboration between teams, management, partners and customers
• Gathers, curates and federates remote institutional and field knowledge
• Provides full live-streaming capabilities without having to know scaling
• On-boards new employees, train and support remote field workers, channel partners and customers

CONSUMER CUSTOMER BUSINESS PROBLEM

The consumer world is ready for mobile live broadcasting. Participation in a live-stream is the next big movement and future of interactive live video.

Facebook serves about 8 billion video views a day. Snapchat serves about 10 billion. Current mobile live-streaming apps do not deliver video in true real-time, failing in typical real world mobile use cases where the existing high bandwidth and battery usage is undesirable.

Consumers also lose their memorable moments as the live-streams are short-lived. GPUcoin will automatically archive these engagements.

e-sporting events constantly draw audiences in the millions, at times matching live football and baseball sporting events. eSports such as Dota2 are also very popular among gamers and live-streaming audiences alike. Being able to stream to millions of viewers with near zero latency and HD quality has been a huge challenge for the existing CDN based live-streaming systems.

Example consumer customers are Dota2 eSports players, free-lance journalists and social-media celebrities who are looking to build their online presence. GPUcoin’s distributed streaming-video network will make this task possible with little to no latency.

SOLUTION TO CONSUMER CUSTOMER PROBLEM / PRODUCT OFFERING

A consumer grade true real-time live-streaming service needs to be built from the ground up to offer live-streaming options for mobile games and eSports on iPhone, iPad, Android smartphones and tablets.

GPUcoin network’s breakthrough open source live-streaming technology brings true real-time video in a scalable way to its audience, with only the remaining network latency. Hoot smart mobile streaming is self-adaptive based on network conditions and available bandwidth and results in significantly lower bandwidth and battery consumption, leading to superior user experience.

Hoot also allows users to stream from their Mac/PC devices, in addition to the mobile apps, directly to engage their social audience. Hoot is the best way to watch, broadcast interactive live-stream videos and discover talented broadcasters.

THE GPUCOIN FOUNDATION AND GOVERNANCE

The GPUCoin Foundation’s primary objective is to promote the real-world application of the GPUCoin Decentralized Open Live-Streaming platform. It also aims to initially develop the GPUCoin platform and advocate governance and transparency for the platform. The GPUCoin Foundation will establish an association consisting of members of the GPUCoin ecosystem, which will be empowered to determine the direction of functionality and improvement to the GPUCoin Distributed GPU accelerate computing, mining platform and associated ecosystem.

GOVERNANCE

As concepts evolve, GPUcoin will work to formalize a governance process built on existing Delegate Proof-of-Stake inputs.

THE DISPUTE RESOLUTION PROCESS

The GPUCoin Foundation will specify a dispute resolution process, utilizing an internationally accepted dispute resolution system. A rotating board of dispute referees will monitor disputes through the resolution process and oversee collateral release to plaintiffs. Note that this board of dispute referees is not the dispute resolution process specifically; rather it is the mechanism through which dispute resolutions can be enacted through the release of collateral on the blockchain.

GPUCOIN TOKEN SALE

The GPUCoin Foundation will fund the development of the GPUCoin Platform discussed in this paper through the issuance of GPUCoin tokens. These tokens will run natively on the Ethereum blockchain and will be offered to backers of the GPUCoin project via a token sale. The token sale will be launched on or about the March 1, 2018. A second token sale will take place once the initial prototype has been developed and tested to fund its deployment.

TOKEN ALLOCATION & DISTRIBUTION

The supply of GPUCoin token is limited to the number of one hundred million (100,000,000) in total (including those available for sale during the Token Sale) & will be generated upon the launch (“Token Launch”.)

The tokens will be distributed in the following manner: 80% (30/30/20) of the tokens will be eventually allocated amongst the community; the remaining 20% will be allocated to the GPUCoin Foundation initiator, early backers, & the GPUCoin protocol network development team.

GPUCoin Token Distribution Model

Channels	Percentage	Lock up Period
30,000,000 GPUCoin token Sale (GTS)	30%	Token Sale – Launch March 1 2018. The initial funding will be used to develop a working prototype, financial setup, legal fees & promotion.
30,000,000 Additional GPUCoin Token Sale (AGTS)	30%	Additional GPUCoin Token Sale. On the release of a successful prototype, a second token sale will be launched to fund the full production ready launch & development of all relevant technology & organization matters.
20,000,000 GPUCoin tokens Retained by the GPUCoin Foundation as Treasury	20%	100% of which locked for 24 months. Strategic Planning, Project Support, Token Swap, Emergency Fund, Development & Legal Fees – These will be subject to a 2-year lock-up. Subsequent to the lock-up, these will be used for various development & operation costs of GPUCoin Platform over 2 further years.
20,000,000 GPUCoin Advisors, Directors and Early Backers	20%	70% of which is locked for 12 months. 30% of which is locked up for 24 months. Distributed to the directors, advisors, and early backers of the project.

[table:hoot_token_distribution_model]

RESTRICTION ON THE USE OF THE FUNDS

To remain in line with the spirit of the project’s open and transparent philosophy, all funds shall be tracked and reported according to the GPUCoin Foundation’s guidelines. A custodian will monitor the usage of the digital tokens and share it with the community periodically.

1. Financial planning and reporting
   1. The GPUCoin Foundation shall develop financial planning and review financial performance of the previous quarter.
2. Digital tokens management
   1. The digital tokens belonging to the GPUCoin Foundation shall be managed by authorized personnel. The security of digital tokens is ensured by multi signature technology.
3. Digital wallet protocol
   1. The GPUCoin Foundation’s digital wallet shall be protected by a multiple signature technology mechanism.
4. Disclosure
   1. On a regular basis, the GPUCoin Foundation shall disclose on the topics regarding community matters, including status of development, operations, and the usage of tokens, as well as whether the GPUCoin Foundation operates in accordance with the governance policy.

TIMELINE

GPUCOIN DEVELOPMENT PROGRESS

CONCLUSION

Over the past several years, advancements in Graphical Processing Unit (GPU) technology and GPU Accelerated Computing have played a significant and growing role in devices such as mobile phones and personal computers as well as in applications ranging from Artificial Intelligence and Machine Learning to 3D Video rendering, oil exploration and various automotive applications.

Every day, GPUs are being applied to a greater range of tasks and applications that can benefit from the chips’ ability to parallel process large blocks of data more efficiently. GPUCoin is the instantiation of a new and innovative decentralized peer-to-peer GPU network. GPUcoin will harness the power of the world’s GPUs for applications that could immediately benefit from the performance that such a network could bring to bear.

Today’s live streaming video systems are complex to manage and difficult to scale, generally requiring proprietary broadcast software to encode captured video, high cost broadcast server software to push video fragments to expensive content delivery networks (CDNs) which cannot deliver the content in real-time. When applying similar architectures to address the needs of Virtual Reality and Video360 applications, the high latency makes these virtually unusable. CDNs suffer from having been conceived in the Web 1.0 wave of Web acceleration advancements, now 15-20 years old. New GPU accelerated video encoding and decoding techniques have made significant strides in more efficiently handling the processing of video and other vector-based graphics.

We are building the layer that will power decentralized, trusted, secure computation. We have a generational opportunity to reinvent computation and usher in a new era for trusted interaction. We believe several next generation Ð-applications will be built on the protocol and incentivized by GPUCoin cryptocurrency. We believe GPUCoin will provide a fertile platform for decentralized cloud distributed computing in trust-less environments and are prepared to dedicate our time and effort to making GPUcoin a reality.

TEAM

Arun Ram

• University of Wisconsin
• Multiple start-ups

Ashok Raj

• Purdue University, MSCS
• Facebook Apps
• iOs App

Bryan Wargo

• Multiple start-up experiences:
  • 2Roam (acquired by Air2Web)
  • AirWave (acquired by Aruba Networks)
  • Nearbuy (acquired by RetailNext)
• Early market entry with mobile, WiFi, data analytics
• Raised over $200M in venture capital and debt financing

---

1: CAGR: Compound Annual Growth Rate / average growth rate per year Source: Statista Digital Market Outlook 2017

[2] https://www.forbes.com/sites/tomcoughlin/2016/08/27/big-storage-is-the-new-reality-in-virtual-reality/#640a6eef4b03

[3] https://datacenterfrontier.com/virtual-reality-data-center/

[4] https://www.networkworld.com/article/2932312/network-storage/5-reasons-you-should-move-your-video-to-specialty-storage.html

[5] https://www.eetimes.com/author.asp?section_id=36&doc_id=1330462

[6] https://www.wowza.com/products

[7] https://www.adobe.com/products/catalog.html

[8] https://aws.amazon.com/cloudfront/

[9] https://azure.microsoft.com/en-us/services/media-services/

[10] New Models for Utility Tokens https://multicoin.capital/2018/02/13/new-models-utility-tokens/

[11] Decentralized IPFS File system https://ipfs.io/

[12] https://github.com/livepeer/wiki/blob/master/WHITEPAPER.md

[13] Ethereum Probabilistic Micropayments https://medium.com/@gustav.simonsson/ethereum-probabilistic-micropayments-ae6e6cd85a06

[14] as AWS and GCP have comparable costs

[15] as the cost of one GPU instance varies between 0.9 – 1.45 $ in both Google compute platform and Amazon web services based on region and demand

[16] https://github.com/ethereum/EIPs/issues/223

[17] http://identity.foundation