DeepSquare - Professional Decentralised Cloud Ecosystem for Web applications, Web3 dApps and High Performance Computing (applying with Spanish DAO)

[DeepSquare]

Name of Project

DeepSquare

Proposal overview

Preamble: This application is done in Collaboration with Harmony Spanish DAO. I personally met recently with Marc Morales while at the Avalanche Summit in Spain. We had a very interesting conversation around current cloud centralisation on how DeepSquare could help. This proposal is the outcome of our discussion. We would be very happy to join the Harmony ecosystem with DeepSquare and this project is not a go/no go for us. Just a way to accelerate DeepSquare becoming available to Harmony users.

We are dedicated to solving the problem of the lack of sustainable, available decentralised compute resources to power Web and Web 3 and Metaverse. We are creating an interoperability layer to link professional computer resource providers (tier2 cloud providers, research centres, …) in a trustless manner. The DeepSquare project wants to tackle all types of compute workloads from High Performance Computing to standard Web hostings. High performance computing powers heavy workload intense disciplines such as AI, edge gaming, simulations, rendering, and is dominated by centralised hyperscalers. Because of the lack of alternatives, most of the blockchain gaming and metaverse projects are forced to use services from the metadonts such as AWS, not because they love them but because they do not have viable alternatives. What DeepSuare is set to do is to change this by creating a decentralised, sustainable computing ecosystem. This inclusive ecosystem provides an opportunity to everyone to build their amazing projects without giving up on their decentralised ethos and enables anyone interested to get involved to the extent they want.

We already have a functional stack and we were able to onboard several HPC users coming from different sectors such as universities, and AI companies. We need additional funding to develop the trustless aspects of the DeepSquare grid (proof of compute blockchain), create a sustainable standard for compute resources and enforce partnership to increase the scope of potential applications that can run on the DeepSquare Grid to meet the growing demand for compute resources within the blockchain (gaming, metaverse, NFTs) but also simply empower project that needs decentralisation.

Where we need help:

• Blockchain / smart contract development: we have a team of developers that build the first sets of smart contracts but we clearly need help on this (at least to review the smart contract code). The development is separated into two tracks: the smart contract layer allows you to purchase compute resources with your metamask. And the trustless grid network will be a PoC (Proof of Compute ) blockchain. PoC can be viewed as a variation of Proof of Stake with extra validation of compute tasks execution correctness.
• Networking/contact with Web3 projects that need to compute resources
• Multi-chain logic to implement an arbitrage mechanism for SQUARE price and allow people to pay for computers from different EVM compatible blockchains.
• We will be hiring researchers experts in consensus algorithm to develop the theoretical framework of what we call Proof of Compute (we currently have discussion with a researcher working with protocol labs, filecoin). We might need help in the implementation of the proof in the blockchain layer.

Milestones Q2:

• Finalise private sale round
• First version of cluster factory (the toolset allowing to deploy clusters both on bare metal hardware and in a virtualized environment). This is the keystone of the DeepSquare grid allowing us to onboard local cloud providers and research centers supercomputers before connecting them to the network to sell spare compute resources.

Milestones Q3:

• Public launch
• Pre-sale of compute resources
• Staking rewards mechanism

Milestones Q4:

• Compute smart contract layer for EVM compatible blockchains (running on testnet) allowing EVM compatible smart contract blockchain to request compute resources in a non-custodial fashion (pay with tokens on their Metamask).

We started our private sale 1 on Ethereum, we now moved to Avalanche to finalise private sale 2 on a much more sustainable blockchain. The current plan is to run the Proof of Compute blockchain as a subnet on the Avalanche ecosystem. But we want to be a multichain project because we strongly believe that the future will be a multi-chain. We want any blockchain users to be able to interact with the DeepSquare grid. This is why we plan to integrate other blockchains after Avalanche. And we decided to choose Harmony as our next blockchain where users will be able to use SQUARE to pay for compute resources. We also want to leverage the Harmony bridges to reach the multichain vision of the DeepSquare project. The vision is to go beyond EVM compatible blockchain and also integrate into the Cosmos ecosystem.

Proposal ask

$300k

There are four axes with potential cross benefits for Harmony and DeepSquare and this is where we focused the current proposal:

1. Our primary focus is on HPC workloads but why limiting our users to HPC workloads. Our software stack is currently already able to handle both standard virtualized workloads (kubernetes based) and specialised HPC workloads (SLURM based). This opens up the possibilities! What we would like to add in our offering is the ability to spin up blockchain nodes for PoS blockchains. We think the decentralisation, professional grade, sustainable aspects of the DeepSquare grid will allow removing the pain point of a single point of failure when running nodes in big tech giants while still giving to the people setting up the nodes unified interfaces to access cloud hoster in many different countries and jurisdictions at the same time. Here a potential collaboration with projects like Pocket Network could clearly be envisioned.
2. The development of the Proof of Compute blockchain will require a Consensus team to build the theoretical proof, we will work with researchers active in the Protocol lab and apply for funding with them. The first version of the Proof of Compute will be based on partial job replication to ensure that actors are providing the requested compute resources to the customers.
3. The development of the Compute Payment Protocol will be done in EVM compatible blockchain. We are looking for smart contracts devs aligned with our vision to complement our current team.
4. The multichain logic implementation

Metrics for success

1. The ability to spin up blockchain nodes using the DeepSquare Cluster factory stack
2. The ability to spin up blockchain nodes in the DeepSquare Grid in many different locations with a single interface.
3. The ability to pay for compute natively in Harmony
4. The first steps towards the multichain vision of DeepSquare

Note : the amount in this proposal does not cover all the costs. But it would already be a very valuable contribution. And also creating a strong link between our emerging community and Harmony.

Success metrics and KPI:

• Successful public token launch
• Onboardingb5-10 tiers 2 cloud provider
• Onboarding 2-5 research centre’s supercomputer
• Onboarding 5-10 startup incubators in the AI space
• Onboarding 5-10 utility companies (for heat re-use)
• Providing DeepSquare branded turnkey sustainable clusters
• Have 2 DeepSquare HPC clusters

External links

Website
Whitepaper

[MarcMorales94]

Wish all the best to DeepSquare! Personally I thinks can really help the decentralised compute resources to power Web and Web 3 and Metaverse due the big hold we are experiencing at this moment.
Let’s take a look, it’s really interesting what this guys are doing… with the possibility to achieve multichain project getting help with bridges.
@Mattyontap

[essalacher]

Thank you for submitting a proposal to Harmony’s $300M Ecosystem Fund. We are in the process of reviewing your proposal and will assign a representative on behalf of the Grants Ops team.

We would also love to have the Harmony community participate to ask questions and provide feedbacks.

If you have more details to add, please do so by replying to this thread.

[Jacksteroo]

@DeepSquare welcome to Harmony!

This is one of my favorite spaces. Cloud infrastructure and node operations. Forgive me for being hyper curious here with some deep dive questions:

• How similar is this with Akash Network? They are ahead by about 2+ years now and it’s a highly complex problem to solve. They’ve found product market fit without focusing on HPC or spinning up nodes for blockchains
• Why build building blockchain be an advantage over writing smart contracts for to perform the accounting and ownership verifications? This could simplify things by first keeping the workload management offchain
• How are you solving the supply chain issue on finding partners to onboard providing the compute workloads? Do you already have partners with data center providers such as Equinix?

For what it’s worth, love the concept of dogfooding node operations on the blockchain. The use case is moot as one could bring down the other and vice-versa. Wouldn’t recommend it

[DeepSquare]

Hi @Jacksteroo thanks for your welcome and interest!

How similar is this with Akash Network? They are ahead by about 2+ years now and it’s a highly complex problem to solve. They’ve found product market fit without focusing on HPC or spinning up nodes for blockchains

There are 3 areas in which we differ from Akash Network.

a) As you correctly pointed out, we come initially from HPC, which allows us to provide resources for highly compute intensive tasks such as machine learning, simulation, etc. Our inference stack (to productionize machine learning models) required the ability to orchestrate container-based workloads with Kubernetes. This opened up a new avenue for us allowing us to generalize the offering to also less compute intensive tasks, and when thinking of the distributed nature of the DeepSquare grid this could be of great interest for software that needs decentralization such as nodes for blockchains. Hence the current proposal!

b) The second aspect is that we focus on professional/enterprise grade clusters, targeting the b2b market. We believe that there should be a certain minimum level of resources available at each location with strong SLAs to allow real businesses to consider this as a true alternative to big cloud providers. In that regard, partners we currently target also include regional/local cloud providers. This also means that we don’t see Akash Network as a competitor. For us there is a world where DeepSquare could provide compute resources to Akash Network. This is similar to the approach we take with OpenFabric with whom we are setting up the partnership.

c) The third aspect is sustainability. For us, it is crucial to make this a priority, and is at the very core of DeepSquare. We develop clusters optimized for the Edge with sustainability in mind. Our current proof of concept is able to reach PuE as low as 0.6 based on the first work from students in Montreal Polytechnique (as part of our ESG workgroup’s initial LCA research). For comparison, standard clouds are well above 1.2 (note: a PuE of 1.0 means that all the energy injected into the computer is used for computing, meaning 0% is needed for cooling). If you are able to reach PuE below one is because the heat generated by the system is injected into district heating systems, as is the case with our PoC running in Sion Switzerland. When talking about grid partners we don’t want to force them to have this level of sustainability, but we want to pave the path to a truly sustainable cloud by creating standards that can be used by anyone.

Why building blockchain be an advantage over writing smart contracts for to perform the accounting and ownership verifications? This could simplify things by first keeping the workload management offchain

Excellent point here! Our blockchain will be a subnet on the Avalanche network. Avalanche subnets are quite flexible and people can pretty much build it as they wish. But as you very correctly pointed out, to make things simpler at the beginning (and also easier to find resources to help developing it) the DeepSquare blockchain we very likely be an optimized EVM based blockchain (WAGMI Subnet Demo | Avalanche Docs) that can already run as a subnet on avalanche, making bridges and many other things much simpler. The first role of our blockchain will be to ensure that jobs requested are correctly carried out by grid partners on the network, and clearly part of the workload management will indeed be off-chain.

How are you solving the supply chain issue on finding partners to onboard providing the compute workloads? Do you already have partners with data center providers such as Equinix?

Very good point. The approach is threefold here.

(1) Utility companies are interested in the sustainable aspects, the heat reuse, and the use of their local green electricity sources. We are in discussion with several of them in different cities in Switzerland to complement our first PoC cluster.

(2) Regional cloud providers/hosters/datacenters: we have built a pipeline of interested parties across Europe and Canada and even have some senior Equinix mgt. in our current investor base :-). The angle taken here is for them to extend their offering with an HPC stack (HPCaaS). The fact also that we reuse spare capacity that they wouldn’t sell otherwise makes their life easier. Because the demand for computing resources varies over time with specific peaks of activities, they very often need to oversize it. And oversizing means spare capacity.

(3) Research centers and university supercomputers: In this case, the open source approach taken by DeepSquare allows them to have a strong and efficient stack for their researchers and again very easily sell spare capacity to the grid.

Concerning your last point which again is great! An important thing to understand is that the way the grid will work will be resilient to the failure of the DeepSquare blockchain. And no running processes will be just hard stopped if the DeepSquare blockchain for some reason experiences a downtime. Failover mechanisms will be put in place to ensure no information is lost on jobs requested and associated user balances using an efficient replicated storage on a trustworthy network, similar to what happens with wardens on the AB bridge (Avalanche Bridge: Secure Cross-Chain Asset Transfers Using Intel SGX | by Conor Leary | Avalanche | Medium)

[Jacksteroo]

Impressive levels of detail.

Building on multiple chains will create an additional layer of complexity, hence adds risk to the project. It’s best that you picked a chain to build with. Harmony is well battle-tested to be built on. Having said that since you’ve picked Avalanche, best for your team to continue building on one chain and not overly complicate matters with cross chain use cases for workload X vs. accounting Y.

Also building for a B2B market is a really long sales cycle, putting this as not a Launch, but an Investment. We’re currently not looking to invest in the HPC field.

We wish you the best of luck in building out this solution. There is a market for this, we’re just not your best partner for this area at this time. Thanks again for considering Harmony.