Detailed Explanation of Akash: A Decentralized Cloud Computing Network

Author: Akash Chinese Community

While the old guardians of cloud computing use more than one server, they can still exercise tyranny over data centers that own a large number of servers. With cloud service providers like Amazon Web Services (AWS), Google Cloud, and Microsoft Azure holding over 60% of the cloud computing market share, most of the internet's data is stored on centralized networks at the behest of just three owners.

The Akash network offers consumers an alternative to centralized cloud providers through its Akash deployment marketplace, allowing users to set the price they are willing to pay to deploy their software, while providers with excess computing capacity bid to host the user's applications. This underutilized computing resource marketplace functions similarly to Airbnb, allowing providers to rent out unused capacity.

The Importance of Cloud Computing

No one can escape the influence of the ubiquitous cloud. Whether it's your phone, car, streaming service, or social media account, you are using cloud technology every day. In its simplest form, cloud computing allows you to access services over the internet rather than on your machine.

In the old days of DVD players, the only legal way to watch Disney movies was to actually purchase the DVD and insert it into a player connected to the TV. Watching many movies required buying a collection of these discs. The cloud changed the mechanism of this physical interaction while retaining the same product: movies. By storing archived movie data on remote servers and running the backend processes for the Disney Plus streaming service, Disney eliminated the requirement for users to own specific product hardware (DVD players), thus enabling it to deliver the same product over the internet.

Disney not only uses the cloud to host customer-facing applications like its streaming service. In terms of business operations, Disney and many other companies use cloud technology for data protection, data analytics, storage backups, server virtualization (virtual machines), and software development. However, companies like Disney do not need to own the server infrastructure that facilitates all these necessary operations. Instead, most companies outsource server management to cloud service providers.

Before the popularity of Amazon Web Services, Google Cloud, and Microsoft Azure, companies owned their own on-premises data centers. From hiring IT teams to manage hardware updates and maintenance to purchasing large real estate and paying hefty energy bills, costs associated with IT infrastructure were recorded as capital expenditures.

However, with cloud providers taking on all costs associated with IT infrastructure, paying for cloud computing is seen as an operating expense, which has tax advantages. Although deploying applications to the three major cloud providers brings financial gains and improves operational efficiency, cloud consumers are locked into the vendor and have limited control over their deployments in a centralized cloud environment.

As more businesses realize the higher levels of scalability, flexibility, and cost savings achievable by outsourcing IT infrastructure to cloud service providers, Gartner predicts that by 2022, end users will spend $482 billion on cloud services. This is a 54% increase from $313 billion in 2020. The pandemic's push for remote hybrid work will only help accelerate the growth of the cloud computing industry.

According to Flexera, 92% of U.S. companies are already using cloud technology, with 90% of them claiming that usage is higher than initially planned due to the pandemic. This contrasts sharply with Europe, where only 42% of businesses were in the cloud in 2021, up from 19% in 2016.

Despite the rapid growth and adoption of cloud computing, 30% of server capacity remains idle across many data centers. As the cloud computing market continues to grow globally, a convergence will occur between vendors seeking to correct supply-side inefficiencies and businesses looking to flexibly scale operations, outsource IT infrastructure, and avoid the complexities caused by centralized entities.

Akash Marketplace

Akash is an open-source platform with a distributed peer-to-peer cloud marketplace that connects users seeking cloud services with infrastructure providers that have excess computing resources. Its platform is used to host and manage deployments and includes cloud management services that utilize Kubernetes to run workloads.

These users are referred to as tenants, typically developers looking to deploy Docker containers to cloud providers that meet specific criteria. Docker containers essentially package code and its dependencies, ensuring that the corresponding application runs the same in any computing environment.

For example, even if the application is developed on a laptop, tested in a sandbox, and sent to the cloud with different configurations, the container can support all three environments without changing the code.

The Akash marketplace operates through a reverse auction model, allowing users to name a price and describe the resources they want to deploy their containers. When cloud providers, from individuals to data centers, have underutilized computing resources, it rents these resources out by bidding to host deployments, much like Airbnb hosts can rent out their extra space.

The cost of deploying containers on Akash is about 10 times lower than any of the three major cloud service providers (AWS, Google Cloud, and Microsoft Azure).

All requests, bids, leases, and settlement payments are recorded on-chain using Akash tokens (AKT).

Akash Leadership

Greg Osuri, Co-founder and CEO: With a background in cloud architecture and entrepreneurship dating back to 2008, Osuri founded Akash as a decentralized alternative to the traditional cloud computing industry. Before Akash, Osuri founded four other companies and worked at prominent firms, including Miracle Software Systems as a technical architect, IBM as a key infrastructure consultant, and Kaiser Permanente as a cloud infrastructure consultant.

Osuri's notable founding projects include AngelHack, one of the world's most diverse developer ecosystems, with over 175,000 developers in 50 cities, and Overclock Labs, a core contributor focused on building a decentralized and open internet.

Adam Bozanich, Co-founder and CTO: Bozanich is a veteran in software engineering, having held senior positions since 2006. With experience across the software development field, Bozanich has worked in QA automation at Symantec, security engineering at Mu Dynamics, and server engineering at Topspin Media. Before Akash, he co-founded two other companies with Osuri: Sproouts Tech and Overclock Labs.

Who Runs the Network, Overclock or Akash?

Overclock Labs essentially guided the Akash network under the leadership of Osuri and Bozanich. Since then, Akash has become a decentralized protocol. Although Overclock still participates in the development of Akash, it does not control the majority of the 100 validators on the network, thus cannot manipulate AKT transactions. Despite the lack of control, Overclock remains a major contributor to Akash's open-source codebase, although many contributors come from outside Overclock.

Since 2020, the number of commits from contributors not affiliated with Overclock Labs on Akash Network's GitHub has increased. A commit is an addition or change to any length of code. Due to the varying complexity and utility of commits, they can serve as an indicator of developer activity within a given project. While Overclock continues to account for most of Akash's development, Akash's own developer community is growing and contributing to the project's development at an increasing pace.

Akash Network

The Akash Network builds its blockchain on the Cosmos SDK framework, utilizing the Tendermint Byzantine Fault Tolerance (BFT) engine for its Delegated Proof of Stake (DPoS) consensus algorithm. To simplify the jargon used in the previous sentence, the core of Akash uses the Tendermint engine as its security and networking layer for its blockchain. It customizes various aspects of its blockchain using the Cosmos SDK, from governance to staking.

Staking AKT in the DPoS Protocol

Since Akash uses a DPoS consensus algorithm, the staking responsibilities between delegators and validators differ. Validators operate nodes to secure on-chain AKT network transactions. While anyone can become a validator, only those with the computing resources and technical expertise to run Akash nodes will.

Akash recommends minimum system requirements for running validator nodes, including a 4-core CPU, 16GB RAM, 256GB SSD storage, and a Linux Ubuntu operating system. Of course, better specifications will equate to better performance; in any case, these are relatively moderate minimum requirements compared to blockchains like Solana.

On Solana, validators require a 12-core/24-thread CPU, at least 128GB of RAM (256GB RAM recommended), and three separate storage units (recommended), totaling 2TB (500GB for accounts, 1TB for the ledger, and 500GB for the operating system).

The frequency at which the consensus algorithm allows validators to approve network transactions—thus earning inflation rewards—is proportional to the amount of AKT held by the validators. This creates a competitive environment among validators to earn AKT either for themselves or from delegators.

Delegators delegate their rights to earn validation rewards on AKT to validators who operate nodes. The validator then allocates rewards to the delegator (proportional to the amount they delegated) and takes a commission on top. AKT holders can delegate through the Keplr desktop wallet.

AKT Tokenomics and Utility

At genesis, 100 million AKT were allocated according to the distribution shown above, with approximately 289 million AKT planned to be released to stakers as inflation emissions, initially set at 100% APR, halving every two years. Governance voting will halve approximately every 3.7 months. According to Akash founder Greg Osuri, the protocol plans to reward cloud providers through emission rewards, but this feature will be implemented later.

Below is a description of the vesting schedule based on the initial allocation of 100 million AKT at genesis (September 2020):

• Investors (34.5% or 34.5 million AKT): Approximately 17.3 million AKT will be released after a one-year lock-up, with about 8.6 million AKT unlocking every six months according to a tiered vesting schedule.
• Team and Advisors (27% or 27 million AKT): 11 million AKT will be released after a one-year lock-up, with about 6 million AKT unlocking every six months according to a tiered vesting schedule.
• Decentralized Cloud Foundation (19.7% or 19.7 million AKT): 1 million AKT will unlock at TGE according to a tiered vesting schedule, approximately 8.9 million AKT will unlock after the first month, followed by approximately 4.1 million AKT unlocking over the next six months, and then approximately 2.1 million AKT unlocking over the next 11 months and two additional six-month periods.
• Ecosystem (8% or 8 million AKT): 2 million AKT will be released after a one-month lock-up, with about 1.5 million AKT unlocking after the first six months according to a tiered vesting schedule, followed by unlocking over the next 11 months and then an additional two months unlocking over six months.
• Testnet (5% or 5 million AKT): 800,000 AKT will be released after a one-month lock-up, with about 1.05 million AKT unlocking after the first six months according to a tiered vesting schedule, followed by unlocking over the next 11 months and then an additional two six-month periods.
• Vendors and Marketing (4% or 4 million AKT): 800,000 AKT will be released after a one-month lock-up, with about 800,000 AKT unlocking after the first six months according to a tiered vesting schedule, followed by unlocking over the next 11 months and then an additional two months unlocking over six months.
• Public Sale (1.8% or 1.8 million AKT): No vesting schedule.

AKT is a utility token with various uses within the protocol, including security, rewards, network governance, and transactions.

Node operators validating transactions on the mainnet and earning rewards must hold a certain amount of AKT, making them among the top 100 holders. This amount comes from the allocation given to validators plus the amount delegated to them. Validators on Akash do not need to stake at least the AKT they hold.

Considering that more AKT increases the likelihood of node operators being selected to validate transactions, thereby increasing their reward frequency. To deter malicious/lazy validators, those who do not adhere to consensus guidelines may have a portion of their stake slashed.

Although not yet implemented, Akash plans to charge a "fee" for each successful lease. It will then send the fees to the Take Income Pool to be distributed to holders. The plan is to charge a 10% fee on AKT transactions and a 20% fee when using other cryptocurrencies. Akash also plans to reward holders for the time they lock their AKT holdings. Holders who maintain their holdings for longer will be eligible for greater rewards.

Only AKT holders can participate in governance. This includes submitting proposals and voting on them. The fee for submitting a proposal is a non-refundable deposit of 1,000 AKT. For proposals that require binary updates (code changes), validators must update the codebase to avoid penalties and continue validating the network.

AKT serves as the reserve currency of the ecosystem, used to pay gas fees and as the default medium of exchange for transactions between vendors and tenants. The Akash whitepaper describes a settlement option that aims to address AKT price volatility in market transactions by locking in the exchange rate between AKT and the settlement currency. This settlement option has not yet been implemented.

Akash's Ongoing Development

After delaying the development and release of the supermini, a home micro supercomputer that can seamlessly integrate with the Akash network for heavy workloads or rental, Akash has learned not to look too far into the future.

With 33 cloud providers supporting 525 active leases at prices approximately 90% lower than traditional cloud computing market prices (due to supply far exceeding demand), Akash is exploring ways to unlock market expansion. It plans to improve its blockchain and cloud marketplace. Development in both areas will make Akash's products more appealing to traditional cloud computing consumers and open protocols for broader crypto adoption.

Akash's testnet 3 was launched on March 7, 2022, indicating the team's focus on developing new features for the mainnet, including:

• Persistent Storage: Workloads containing large datasets (like blockchain nodes) will be able to persist data between restarts. If a deployment goes down or is reinitialized, data will not be lost.
• Fractional uAKT: Eliminating the implicit minimum deployment cost (one uAKT per block). Fractional uAKT will make pricing more consistent and resource consumption more accurately reflect costs.
• AuthZ: Users can authorize one wallet to spend a certain amount of tokens on deployments through another source wallet. This can reduce security issues when large teams deploy together without using a large shared wallet. It allows new community members to initiate deployments without using a faucet.
• Inflation Curve: An automatic curve mechanism that corrects itself when inflation deviates from the plan outlined in the whitepaper. This eliminates the need for manual intervention that requires quickly submitting governance proposals to implement fixes.

A complete list of Akash's plans for development in 2022 can be found on its roadmap. According to Akash founder Greg Osuri, most deliverables for Q4 2021 have been developed and are still in the process of being put into production.

Akash's target market, from Web3 to artificial intelligence and machine learning (AI/ML), continues to grow in parallel. As the Web3 crowd seeks decentralized computing infrastructure, and the heavy computing crowd seeks low-cost flexible solutions to deploy software, Akash will have many opportunities to meet the demands of the burgeoning cloud computing user base.

Decentralized Cloud Competition

Akash is not the only blockchain project disrupting the nearly $5 trillion cloud computing market. Other decentralized competitors include Dfinity (Internet Computer), Ankr, and Cudos. While Ankr launched with a model similar to Akash, today, Internet Computer and Ankr approach cloud computing from different angles. Meanwhile, Cudos has nearly replicated Akash.

Dfinity / Internet Computer

While the Internet Computer is primarily regarded as a smart contract platform, it attempts to decentralize everything within its protocol, from the consensus algorithm to the servers that validator nodes connect to when approving network transactions broadcast over the Internet.

Claiming decentralization of the consensus algorithm is not unique. Almost every project claims it, but few truly adhere to it (look at Solana—validators cannot plan to restart the network on a decentralized system). However, the Internet Computer also claims to decentralize the servers used by its validator nodes to execute consensus. This contrasts sharply with Ethereum, where about 25% of nodes ran on centralized AWS infrastructure in 2019.

The Internet Computer is not a direct competitor to Akash's underutilized computing resource marketplace. Instead, it creates a system where independent node operators pay data centers to host their nodes, while cloud providers are compensated for the computing power contributed to facilitate network activity.

The Internet Computer attempts to let a truly decentralized protocol run on decentralized data center networks by pairing the computing resources provided by globally independent data centers with its node operators, rather than servers owned by AWS. Although it similarly utilizes excess computing capacity, the Internet Computer does not generate market activity for cloud computing like Akash does. It may capture a share of potential cloud providers but does not pose a significant threat to Akash's market.

Ankr

The Ankr project was launched with the goal of utilizing idle computing resources from data centers to build an open-source cloud, referring to it as a "distributed cloud computing network." This would make Ankr an almost exact replica and direct competitor to Akash. However, it does not allow users to run any applications on the cloud services proposed by Ankr but instead focuses on providing access to Web3 infrastructure specifically through RPC (Remote Procedure Call) used to make requests to server processes.

Focusing on Web3, node providers join the Ankr network to offer cloud services, connecting those building decentralized applications, protocols, and smart contracts to various blockchains through API endpoints. Ankr serves over 50 PoS blockchains using its global node delivery system and developer toolkit. Similar to the Internet Computer, nodes supporting its RPC layer in the Ankr ecosystem may only threaten the number of providers Akash can connect with. Again, this does not pose a significant threat to Akash's market.

Cudos

Cudos is another project within the product line offered by Cudo Network. Using Cudo Network's Cudo Miner product (distinct from Cudos), any user wishing to mine must download the Cudo Miner application, select the token to mine, and the application will mine the selected token when the machine is idle.

In the same spirit of utilizing idle computing capacity, Cudos will offer a product very similar to Akash, connecting underutilized cloud providers with builders looking to deploy WASM containers and virtual machines.

As of this writing, Cudos is building its mainnet using the Cosmos SDK, with its testnet accessible via the Big Dipper block explorer. Although it connects idle computing capacity with builders, Cudos does not plan to operate its marketplace through reverse auctions, nor does it plan to open-source its codebase like Akash. With its mainnet yet to launch, Cudos is far behind Akash in capturing market share from both supply and demand sides of the cloud marketplace.

Virtual Machines… Maybe Later

Virtual machines (VMs) are an extension of the old guard, a way to implement virtualization solutions in the past, and Akash has no explicit plans to support them. In a conversation with Akash founder Greg Osuri, he mentioned that computing providers have no restrictions on offering VM support—even expressing interest—but at this stage of Akash's development, it doesn't make much sense. As an evolving protocol, Akash focuses on supporting the technology through an ever-expanding user base rather than allowing enterprises to use it as an excuse to avoid change.

While some industries still question the ability of containers to handle heavy applications like AI and machine learning in a manner similar to VMs, research shows that even if container performance is not better, it is negligible. Akash recognizes that the demand for cloud computing for AI/ML is accelerating and is prepared to address potential growth beyond its typical Web3 tenants.

Praise, Not Detriment

Akash is not looking to replace the centralized cloud computing industry but rather to complement it. If the conspirators are correct, Ethereum could indeed be killed by Solana, Polkadot, Cosmos, NEAR, or Algorand (not really, but long-term Algonauts can dream). The fit of cloud computing and Akash in the industry is different. Just as Airbnb did not destroy hotels, Akash will not destroy centralized data providers. Both Airbnb and Akash simply allow individuals to rent out unused capacity.

Akash may be able to attract cloud customers by offering prices at a fraction of the cost of the three major providers, along with flexibility and control over the resources used for hosting deployments without vendor lock-in. As traditional enterprises establish crypto departments and become more culturally attuned to Web3 technologies, Akash will be at the forefront of decentralized cloud computing with a functioning mainnet and a business model that mitigates supply-side inefficiencies.