In an era where artificial intelligence is reshaping our technological landscape, it is imperative to reconsider how we design and operate data centers. With decades of experience in the Internet and Cloud space and in response to technological evolutions, ThreeFold is proposing a new approach.

Over the last 30 years, data center construction has largely remained static. Contemporary data centers risk becoming obsolete if they continue to neglect evolving needs. As we stand at this technological crossroads, the urgency to adapt is clear: unless we pay close attention, new data centers risk repeating the inefficiencies of their predecessors, leading to a poor utilization of investments and a significant mismatch with current and future requirements.

🔗The Two Main Flaws of Existing Data Centers

🔗Inadequate Power Provision

The advent of AI has dramatically escalated power needs. An AI-centric rack can consume up to 200 kW – a staggering 15 times more than the capacity of data centers built in the conventional model. This exponential increase poses significant challenges, particularly when dissipation exceeds 50 kW per rack. Traditional designs fall short in crucial aspects like cooling and power distribution, with conventional air cooling methods becoming increasingly unviable.

🔗The Shift in Market Needs: Beyond Co-location

The market dynamics have evolved; selling space in a data center, akin to ‘advanced real estate’, is a model that is rapidly losing relevance. The traditional co-location framework, which expects customers to manage their hardware and infrastructure within leased rack space, is becoming antiquated.

In the modern landscape, the role of a data center should transcend mere space rental. The emphasis should be on providing integrated solutions – selling compute, storage, and network capacity, not just rackspace and power. It’s about delivering a comprehensive service that aligns with the complex, ever-evolving needs of businesses immersed in the digital and AI revolution.

🔗The Data Center Landscape Today

Tier 3 and Tier 4 Data Centers are classifications outlined by the Uptime Institute, which provide standards for the design, construction, and operation of data centers. The Institute defines four tiers (1, 2, 3, & 4) to categorize data centers according to their design and infrastructure capabilities. For the purposes of this article, we will focus on the top two tiers.

🔗Tier 3

A Tier 3 Data Center is a type of data center facility that is designed to provide a high level of availability and redundancy for IT infrastructure and data storage.

🔗Key Characteristics

• Availability: Tier 3 Data Centers typically offer 99.982% availability, translating to around 1.6 hours of downtime per year.
• Redundancy: They feature N+1 redundancy for power and cooling systems, providing backup components to ensure continuous operation in case of failure.
• Concurrent Maintenance: Tier 3 Data Centers allow for maintenance and repairs to be performed on systems without disrupting operations.
• Cost: Tier 3 Data Centers are less expensive to build and operate compared to Tier 4 due to lower redundancy requirements.

It's important to note that while Tier 3 Data Centers provide a high level of availability and redundancy, they are not as robust as Tier 4 Data Centers, which are designed to provide even higher levels of fault tolerance and uptime. The choice of data center tier depends on the specific needs and risk tolerance of the organization using the facility.

🔗Key Challenges

• Limited Uptime: Despite high reliability, Tier 3 Data Centers may still experience noticeable downtime, which could impact critical operations.
• Single Points of Failure: While designed for redundancy, Tier 3 Data Centers may still have single points of failure, potentially leading to downtime during component failures.
• Scalability: They may have limitations in scalability compared to higher tier data centers, potentially requiring more significant upgrades for expansion.

When a business is considering colocation or hosting in a Tier 3 Data Center, it's important to assess its power requirements carefully. For operations with high power demands, it might be necessary to discuss with the data center provider the possibilities of allocating additional power resources or choosing a high-density zone within the data center. Data center providers typically measure power in kilowatts (kW) per rack and will have different pricing models based on the power and cooling requirements of the equipment hosted.

🔗Tier 4

A Tier 4 Data Center is the most advanced type of data center tier, as classified by the Uptime Institute. The Tier 4 Data Center is the most robust and less prone to failures. Its design ensures redundancy and reliability.

🔗Key Characteristics

• Availability: Tier 4 Data Centers offer the highest level of availability, typically guaranteeing 99.995% uptime or just around 26.3 minutes of downtime per year.
• Fault Tolerance: They feature a fault-tolerant design with 2N+1 redundancy for critical systems, ensuring uninterrupted operation even during component failures.
• Redundancy: Tier 4 Data Centers have redundant power and cooling distribution paths, eliminating single points of failure.
• Security: They often have advanced security measures in place, including biometric access controls and surveillance systems.

🔗Key Challenges

• Complexity: Tier 4 Data Centers are more complex and expensive to design, build, and maintain due to their fault-tolerant architecture.
• High Cost: The increased redundancy and fault tolerance come with higher costs, making Tier 4 Data Centers more expensive to operate compared to lower tiers.
• Overkill for Some Applications: For organizations with less critical operations or smaller budgets, Tier 4 Data Centers may be overkill and not cost-effective.

🔗Choosing Between Tier 3 and Tier 4

• Cost vs. Need: Tier 4 Data Centers are significantly more expensive to build and operate. Organizations should assess whether they need the highest level of uptime and fault tolerance that Tier 4 provides.
• Business Requirements: Businesses that can afford occasional downtimes for maintenance or those that don't have extremely critical data operations may find Tier 3 facilities adequately sufficient.
• Industry Compliance: Certain industries may require or favor Tier 4 facilities due to regulatory requirements.

In summary, while both Tier 3 and Tier 4 Data Centers are designed to ensure high levels of uptime and reliability, Tier 4 takes this a step further with increased redundancy, fault tolerance, and a correspondingly higher price point.

🔗Introducing the Tier S Data Center

The Tier S model is designed to exceed the capabilities of traditional Tier 3 and 4 Data Centers, addressing the increasing demand for infrastructure that can support emerging IT workloads with greater efficiency and resilience. Where "S" stands for "Self-Healing," Tier S Data Centers present a new way forward, while remaining backwards compatible.

The Tier S "circular approach" allows us to build the most reliable and redundant data centers in the world. A larger data center (typically 2-10 mW capacity) is extended with at least nine edge data centers. Key features include advanced liquid cooling technology and cybersecurity mechanisms, enhancing the center's ability to handle complex IT workloads like AI and ML.

Flexibility is central to the design, with each edge data center capable of housing 3-30 pods (subsections), and a versatile main data center. The innovative self-healing technology ensures continuous, secure operations, reducing downtime and maintenance needs. The infrastructure, integrating state-of-the-art hydro coolant distribution units, is designed for high-performance computing, meeting emerging technological demands.

🔗Key Benefits

• Superior Reliability: Data storage is highly secure and can't be lost, would even be safe if quantum computing would be used for hacking.
• 100% Redundancy: Fully redundant power and network paths. All hardware is equipped with liquid cooling for enhanced reliability and energy-efficiency..
• Capacity and Expansion: A main data center with a 2-10 MW capacity, supplemented by at least 9 edge data centers.
• Flexible Configuration: Each edge data center can include 3 to 30 pods (more information below), with the main data centers configurable as pod, rack, or container-based.
• Disaster Resilience: The circular design, with a main data center and edge data centers, offers robust protection against natural disasters, war, or unrest.
• Self-Healing: Equipped with self-healing properties, Tier S Data Centers ensure that workloads can operate safely with minimal need for human intervention.

🔗In Conclusion

The existing data center model is simply insufficient for emerging demands of the IT world. Billions of dollars will be mis-spent if we continue down the existing path. This is why ThreeFold is presenting a new approach. While the Tier S Data Center is still in its conceptual phase, we are actively working with partners on implementing the first of its kind in the near future.

DePIN Hub is "the ultimate resource to learn more about the DePIN ecosystem," created by the Hotspotty team and launched officially yesterday. (Congrats!)

Take a look at the ThreeFold project listing and feel free to let us know if you have any feedback or share within your networks!

DePIN stands for Decentralized Physical Infrastructure Networks, and covers projects ranging from connectivity to clean energy production to projects like ours.

In today's digital landscape, the development of a decentralized open-source Internet infrastructure marks a significant leap forward. However, while the foundational structure of this innovation brings promise, its inherent reliability isn't guaranteed. After more than five years of building and expanding TF9's open-source technologies, it's now time we pivot towards commercialization. Ensuring a robust, dependable, and trustworthy offering becomes paramount. In this post, we'll share the various ways that reliability is being integrated with TF9 technology.

🔗Technical Redundancy

Our quantum safe storage system, coupled with replication, and web gateways, presents a pathway to achieve technical redundancy already. While these capabilities exist, we admit that at this time it is too complex for a starting user. Making this easier through refined tooling and a strong user experience will be imperative in order to allow a wide audience to take advantage here.

🔗TF9 Cloud

Our upcoming commercial offering, TF9 Cloud, will bolster reliability and trustworthiness. Through TF Cloud, we will need to provide comprehensive service, robust support, service level agreements (SLAs), and streamlined payment methods, including credit card options. The development of user-friendly tooling will significantly simplify and easen the process as well, enabling users to deploy and maintain reliable workloads effortlessly.

🔗Farming Pools

The introduction of farming pools in TF Grid 4.x will play an important role in fostering reliability as well. These pools not only contribute to resource pooling but also offer visibility into the network's participants. This transparency enhances trust and aids in identifying reliable contributors within a decentralized framework.

🔗Establishing Reputation and Transparency

Building a system that emphasizes reputation and complete visibility regarding service levels and uptimes is pivotal. By offering insights into each participant's performance metrics, Grid users will be able to more effectively choose a trusted provider.

In essence, through a combination of technical and strategic advancements, our journey towards a dependable and trustworthy decentralized Internet infrastructure is underway, setting the stage for the next phase of ThreeFold.

Imagine this: two friends in East Africa trying to connect via a video call. Instead of a direct link, their data embarks on a round trip to far-off data centers, consuming resources and adding costs. It's a snippet of a larger issue—how centralized infrastructure is holding back progress in connectivity.

🔗Breaking the Distance Barrier

In many regions, the internet economy remains anchored in distant lands due to reliance on long cables to Europe, North America, or China, where colossal data centers reside. This scenario keeps data, jobs, and economic benefits outside developing regions. But what if connectivity was local, decentralized, and efficient?

🔗The East Africa Example

Picture our technology in action. Instead of relying on data centers thousands of miles away, our decentralized infrastructure brings the connection closer. Those friends in East Africa can communicate peer-to-peer, reducing bandwidth needs by up to 10 times while boosting security. This isn't just about video calls; it's about revolutionizing every digital interaction.

🔗Streamlining Complexity

The tech world often gets tangled in layers of abstraction, favoring integration over innovation. While layers aid compatibility and incremental progress, they hinder true change. Excessive abstraction results in inefficiencies, performance lags, increased costs, and scalability woes.

🔗The Education Gap

Innovation takes a back seat when education leans heavily on commercial technologies. The push for quick-fix solutions neglects long-term innovation. It's like opting for a painkiller instead of addressing the root cause of the problem.

🔗Local, Secure, Reliable

Our vision? A decentralized internet where compute, storage, and network are local, sovereign, and secure. Storage becomes more reliable with less overhead, doubling up as a global content delivery network. The network stays resilient, routes around issues, encrypts end-to-end, and knows precisely where data originates and ends.

🔗Tackling the Root Issues

By constructing a decentralized, autonomous internet infrastructure from scratch, we're tackling these issues head-on. Minimizing complexity while solving problems at their core is our mantra. It's not just about a tech revolution; it's about reshaping connectivity to be inclusive, efficient, and empowering for everyone, everywhere.

The digital world stands at a crossroads. We can either perpetuate the status quo, or we can pave a new path—one that's decentralized, efficient, and equitable. The time for a decentralized internet that solves problems at their roots is now. It's time for connectivity that knows no boundaries.

The power to transform connectivity is in our hands. Let's reimagine the internet not as a distant entity but as a network that thrives locally, serves globally, and empowers universally. Together, let's build a decentralized internet that propels us into a future where connectivity is a right, not a privilege.

With new devices being connected every day and emerging technologies like the Internet of Things (IoT), the metaverse and Web 3 gaining popularity, the demand for Internet capacity is booming. Already, these developments are shifting the Internet’s “center of gravity” from its core to the edges in order to keep up with the growing demand. In an increasingly interconnected world, centralized cloud architectures are proving to be bottlenecks hindering future growth.

Here are some interesting numbers to emphasize the explosive growth of connected devices and the impending data explosion:

• Back in 2018, 90% of global data creation took place within the two previous years.
• In 2022, 4.95 billion people are using the Internet and about 2.5 quintillion bytes of data are created on a daily basis.
• About 2 billion computers, 15 billion mobile devices and more than 22 billion IoT devices are active worldwide on a daily basis.
• By 2025, more than 1 billion people and 75 billion IoT devices are expected to join the Internet.

And yet, the expansion of Internet capacity still largely occurs in the same centralized and inefficient way. The client-server model of hyperscale data centers is unable to meet this rapidly growing demand due to high building and operational costs to fulfill bandwidth, performance and cybersecurity requirements. As data centers are primarily built in the US, China and Europe, emerging countries have to be connected via sea cables which makes the model’s limitations particularly severe for these countries. And even in countries where many data centers are located, they are built in regions where land and electricity are as cheap as possible. Consequently, the bulk of data centers is still far away from where the majority of users and connected devices reside, and where huge amounts of data are generated that require processing – resulting in a weak distribution of Internet capacity.

About 175 Zettabytes of data will be created by 2025, as our economy becomes increasingly digital and data driven. And this amount of data needs to be captured, stored and activated in a secure, efficient and high-performing way. So, the aforementioned technological use cases will not only accelerate the demand for Internet capacity but also require edge capabilities, such as low latencies, in order to run and scale successfully. In fact, the centralized data center model poses one of the biggest obstacles to the success and spread of the Internet of Things and related applications. This will have to change. Actually, industry experts expect 75% of data to be processed outside of the traditional data center or cloud model by 2025.

🔗The rise of edge computing

Computing needs to be pushed from centralized data centers to the edge to support the massive data consumption coupled to the rise of connected devices and emerging technologies. For instance, as more and more home appliances and devices are getting interconnected through the IoT economy, the amount of real-time data generated within our homes on a daily basis is rapidly expanding. Edge computing tackles the shortcomings of the centralized model in terms of bandwidth, latency and autonomy. It enables applications and solutions to be hosted closer to where their data is being produced and consumed, thereby allowing for low latencies, rapid data processing and real-time decisions. 

The industry analyst company IDC estimates the global spending on edge computing to increase by 14.8% compared to 2021 – amounting to roughly $176 billion in 2022. By 2025, this number is predicted to reach about $274 billion.

The idea of edge computing’s ability to process data on-site is simple enough. Nevertheless, edge computing fails to deliver on its promise as it still remains somewhat centralized today. Granted, so-called edge data centers are closer to where the data is being consumed than hyperscale data centers. But while they do provide an improvement from traditional data centers, they are not able to provide far-edge capabilities, i.e. moving as close to the user and point of data creation as possible. They are simply smaller types of data centers that are still operated in a centralized manner, require the same infrastructure elements as centralized data centers, and usually rely on bigger, central data centers. Quite edgy isn’t it?

As the need for Internet capacity – specifically at the edge – is rapidly intensifying, a new approach to Internet infrastructure is clearly needed. The Web 3 movement created a spur of interest across industries due to its promise to decentralize the Internet and cloud as we know it. Future data will be generated at the edge of the network and Internet which certainly requires an architectural shift away from the current model.

🔗Realizing an interconnected future

Ironically, decades after the spark of the open-source peer-to-peer Internet movement was steamrolled by large tech companies and governments eager to centralize the global Internet, we seem to enter an era of decentralization again. There’s no doubt that large tech companies did bring a lot of innovation to our global Internet and cloud stack. Technically though, peer-to-peer systems have always been the most scalable option. Moreover, open-source architectures have built a strong reputation for providing high levels of security and flexibility. 

In a peer-to-peer model, people and devices are able to exchange directly, without intermediaries or servers in the middle – leading to a more energy-efficient, scalable and flexible paradigm that can expand on-demand. And undoubtedly, open-source will play a key role in edge computing by simplifying and enhancing the building and deployment of software at the edge.

By leveraging open-source and peer-to-peer technologies, TF9 has come up with an Internet infrastructure that is capable of far-edge computing. TF9 offers a powerful set of compute primitives that provide the agility to create and deploy containers, Virtual Machines (VMs) and Kubernetes environments at the edge. Having integrated all leading industry-standard tools, ThreeFold’s solution allows for a seamless migration of any digital workload, making TF9 the ideal infrastructure for edge workloads. And all data produced by these workloads can be stored in a way that is ultra safe, highly reliable and unbreakable. Combined with end-to-end encrypted peer-to-peer networking, applications can be hosted closer to where the data is being consumed – representing the ideal underlying layer for IoT networks, blockchain-based workloads and more to run and scale at the edge seamlessly, efficiently and privately.

The shift towards the edge is inevitable. TF9 is the first full-stack, decentralized, peer-to-peer Internet infrastructure in the world that combines compute in the form of highly-optimized containers, storage and network capacity in an all-in-one solution. With  more than 81 Petabytes of capacity distributed across thousands of 3Nodes and 76 countries as well as plans to sponsor the expansion of the Internet to developing regions, TF9 is the foundation for a thriving interconnected future that is secure, borderless and at the edge.

Today, the global Internet and IT infrastructure requires enormous amounts of energy, responsible for about 10% of annual global energy consumption. This makes the IT industry amongst the most pollutive industries in the world (along with the often-discussed airline industry, for comparison).

We believe IT can be sustainable. Actually, it should be! That's why energy efficiency has always been a key focus of ours. With climate change and sustainability as some of the most pressing issues of our times, new green solutions need to be our priority. It has been key to ThreeFold since the very beginning of its story to enable a technological infrastructure that can reduce the Internet’s carbon footprint drastically.

🔗Simplicity is the key to unlock a better future

Over time, integration suites, middleware solutions, and enterprise service busses have been invented and implemented to cover integration challenges within the current internet. This has further complicated IT architectures, resulting in a loss of actual end user workload performance.

ThreeFold reworked the whole approach to IT architectures and eliminated the layers of complexity. The result is a cloud stack that presents minimal overhead and that requires less hardware – reducing heat generation and eventually cooling requirements. By reducing the need for power, the net result is an energy-efficient Internet grid.

🔗Improving resilience and efficiency through Self-Healing IT

Keeping systems up and running is a major requirement in modern day IT. Leading vendors earn most of their margin by selling maintenance contracts, performance guarantees, and professional services. Their business models bring no incentive to make things simple and efficient for consumers.

Creating a self-healing environment requires only minimal manual intervention to replace broken hardware components. Broken hardware components are inevitable, and the Zero-OS distributed node architecture deals with such failures by turning off broken components and provisioning unused capacity. By removing the need to ship broken parts back and forth, and linking to the corresponding installation knowledge (engineers), a lower carbon footprint can be achieved.

🔗“No painkiller” approach. We tackle the root issue

If a specific piece of software or hardware is not delivering the required performance or reliability, we should analyze the core design/algorithm and its usage of soft/hard components to determine the root causes of persistent issues to be solved.

Currently, most storage vendors follow the path of least resistance when trying to improve on performance. Their solution for meeting performance targets is to use faster components (CPU, memory, network card, proprietary acceleration) rather than looking at the core algorithms that drive the utilization of those components. Swapping components for faster ones perpetuates a cycle of cutting corners and focuses reliance on innovation from hardware component providers rather than solving the more fundamental and underlying issues. That's a "painkiller" approach.

Zero-OS was designed to solve root problem issues by design. While still using high-performance hardware components on the ThreeFold Grid, the algorithm of Zero-OS maximizes utilization possibilities. This enables greater performance and sustainability. Some key examples how this is achieved:

🔗Eliminating unnecessary elements from standard architectures

Virtualization adds a layer of software between the actual end-user workload and physical hardware. It also allows for multiple workloads to run on the same hardware by using excess capacity. There is definitely merit in building a virtualization solution. Building an effective virtualization solution that does not require abundant context switching is key and this is why Zero-OS has been developed. Zero-OS uses a minimal Linux kernel that allows for a number of user spaces to co-exist. In these user spaces, containerized versions of software can be run, eliminating the need for hypervisors, virtual OS to fuel the virtual machines, and guest operating systems. This also minimizes the required overhead for the host OS.

🔗Minimizing network connections usage

Supercomputing delivered many new technologies. But, not all of them are usable solutions for everyday workloads. One technology invented to make supercomputers performant is the use of Remote Direct Memory Access (RDMA). RDMA allowed physical CPU boards (containing both CPUs and memory) to access other CPU boards in memory stored data over a dedicated channel. This eliminated the need to transport data between CPU nodes over network connections (or other mainstream means to exchange data). The result was a considerable reduction in overhead created when allowing distributed end user workloads to operate over multiple physical cores. The leading server and storage solutions brands have never considered nor implemented such mechanisms. This resulted in an increased need for faster networks - increasing complexity, costs, efforts and resources needed to operate solutions.

RDMA is only used in certified farms and requires involvement of TFTech.

🔗Minimizing disk usage

The storage solution uses slow and big (the bigger the better) HDD disks which drive in rack density. As they spin slower they consume less power and need less cooling. But then, how can you deal with read/write intensive workloads? The storage algorithm uses a SSD cache to acknowledge Inputs/Outputs coming from the applications, fills the erasure coded data blocks in large (up to 64MB) storage containers and writes the big chunks of data on the big and slow HHDs. If you are familiar with HDD disks, you know they like to be streamed on instead of having a lot of small bits written in a scratchy way.

Continuous innovation in all of these areas are leading to an efficient technology stack.