Compute Power is Compacting and Moving to the Edge

The last few years have seen a quantum leap forward in technological progress, driving demand for unprecedented speed, bandwidth and processing power. A new wave of entrepreneurship is flourishing, with enterprises pursuing technologies and business models that are dependent on ultra-low latency data transmission.

Industries such as healthcare, finance, energy, media, engineering, retail, manufacturing and logistics are leveraging technologies that are too reliant on high-speed networking to run on traditional, centralized cloud servers. These next-generation applications include autonomous vehicles, interactive sports entertainment, smart video surveillance, industrial automation with machine vision and digital energy mining, just to name a few. Furthermore, the proliferation of mobile devices and a new work-from-anywhere dynamic created by the Covid-19 pandemic requires seamless, always-on connectivity across a much more distributed population. 

Edge computing, where data is processed on site or in devices rather than at a centralized server, limits the fiber distance between the data source and the end user, greatly reducing latency while increasing capacity and reliability. Furthermore, locating dense compute closer to the edge can drastically reduce backhaul charges—the costs associated with communicating across the global internet.  

For enterprises dependent on ultra-low latency network performance to enable business execution, edge computing is a critical component of their IT infrastructure. Gartner even predicts that by 2025, a whopping 75% of enterprise-generated data will be “created and processed outside a traditional centralized data center or cloud.” 

While the shift to the edge is necessary to process massive data sets and enable the use of these next-gen technologies and applications, edge computing requires something that is often in short supply: space.

Limited Real Estate = Limited Processing Power

A significant amount of data center space is taken up by cooling systems, the most common form being air cooling. Air cooling not only requires bulky computer room air conditioning (CRAC) and computer room air handling (CRAH) units, but it can also require large ductwork or raised floors (aka, “access flooring”) up to 1 meter high, just to supply air where it’s needed most or provide a path for air to return back to the cooling system.

This may not be a major issue for data centers with plenty of land to build on, but edge data centers are typically built in urban areas, where space is limited and far more expensive. Edge data centers need to be built close to dense user populations to service as many businesses and customers as possible. Urban locations are not only close to large numbers of users, but are also close to long-haul fiber and backbone networks.

The Core Dilemma 

Edge data centers must support dense servers and heavy workloads in compact spaces. Air cooling and other traditional cooling methods do not allow for this compaction, because the machines, heat sinks, air ducts, raised floors and other parts of the cooling infrastructure take up too much space. Additionally, air cooling is so noisy that in many locations it’s a non-starter because of local noise regulations.

The solution? A cooling system that not only allows for high thermal thresholds of compute power and massive data processing, but also is silent and takes up significantly less space. That solution is 2-phase Immersion Liquid Cooling.

2-Phase Immersion Cooling is the Perfect Fit

2-phase immersion cooling is a groundbreaking technology that provides the necessary cooling capability to enable edge computing and massive data processing in any environment, even harsh outdoor locations, while drastically reducing the need for space. 

2-phase immersion cooling involves submerging servers in compact, self-contained DataTanks™. The servers are immersed in a dielectric (non-conductive) fluid, which absorbs the heat generated by the servers until it reaches its boiling point. The transition from liquid to gas is called a phase change, hence “phase one” of the process. 

The second phase change in 2-phase liquid immersion cooling is that of the gas changing back into a liquid, thanks to a condensing coil placed just above the fluid. Condensed fluid collects on the coil then falls back into the tank, where the cycle is repeated continuously. No additional coolant pumps are required because the process is self-contained, and the fluid itself rarely has to be maintained or replaced. 

This process is entirely passive and there is virtually no power consumed by the 2-phase liquid immersion cooling process inside the DataTank. A low-wattage programmable logic controller, a color touchscreen display and a fluid filter pump are the only items that consume power, making 2-phase immersion the most efficient form of liquid cooling.

Not only does this eliminate the need for large, bulky cooling systems, but thanks to the efficiency of this method, the hardware – specifically the CPU boards – can be densely packed, with mere millimeters of space between boards. This drastically reduces the actual space required for the IT itself by up to 82% by volume.

To put this in another perspective, a 36MW air-cooled data center would effectively occupy over 200,000 square feet, whereas a 2-phase liquid immersion-cooled facility would occupy just shy of 88,000 square feet. This represents a 56% savings in physical building footprint and a 69% reduction in whitespace. With DataTanks handling 150kW of compute power, the IT in 3,600 racks can be housed in just 240 2-phase immersion cooled tanks. And even greater compaction can be achieved by using the full 252kW of heat rejection per DataTank.

Enabling powerful edge data centers to be built in small, compact urban spaces is a game changer in itself. But 2-phase immersion cooling also enables you to bring the power of the edge to virtually anywhere in the world.

2-Phase Immersion Cooling Can Bring the Edge to You 

Imagine a 100kW data center in a 10×10 container with world-class computing performance. According to Microsoft, 2-phase liquid immersion techniques constantly yield CPU clock rates up to 51% better than air-cooled solutions and 24% better than direct-to-chip cold plates. These advancements in performance wouldn’t be possible with the thermal limitations of air cooling; the over-clocking of servers would exceed thermal boundaries. Until now.

The proliferation of “micro” data centers, combined with 2-phase immersion cooling, means the power of the edge can be brought to wherever the data is being generated, not just cities. 

For example, a rural hospital can now leverage the power of edge computing for AI, advanced analytics, real-time data relay between medical monitoring devices, and many other technologies that provide better and faster patient care, especially in critical cases. Hospitals no longer need to connect to a regional or centralized cloud data center to gain the processing power they need; a micro data center can be placed in or just outside the hospital, cooled by 2-phase immersion liquid DataTanks.

Or imagine an energy exploration vessel mapping the ocean floor with the most advanced equipment to find oil and gas. The surveying process used to take up to six months. Massive troves of data would be collected and stored on the ship, then taken to advanced data processing labs for analysis after the ship returned to port. It would then take another half a year for the exploration report to be completed. Now, with micro edge data centers, the processing and analysis can be done on board the ship as soon as the data is collected. 

This new paradigm opens a world of possibilities for many industries and organizations. Micro data centers with sealed DataTanks can operate in the harshest environments, where dust, dirt, debris, sand, salt, particulates, and other contaminants would typically hinder or prevent air cooling. Energy exploration, digital oil fields, mining, distribution and logistics, automated manufacturing, critical healthcare, financial markets and many other applications can now process troves of data on site and perform analysis in real time, breaking through previous constraints and expediting time to market.

The Edge is Now Wicked Cool

The need for regional and micro edge data centers, located as close as possible to where data is generated, is growing by leaps and bounds. However, this need can only be fulfilled by reducing the physical footprint and ecological impact of the data center itself.

Key Stats of 2-phase Immersion Liquid Cooling:

  • +30X more heat rejection than air-cooling (up to 252kW per 48RU)
  • Nominal pPUE of 1.02 to 1.04 (site-dependent)
  • 69% reduction in whitespace (based on 150kW per DataTank)
  • Zero water consumed for heat rejection
  • No noise emission from DataTank
  • Future-proof: DataTanks are forward compatible for almost all types of racks or IT hardware
  • USD $3.5M reduction in data center build cost per MW (immediate payback) vs. air-cooling

2-phase immersion cooling provides the best efficiency of any cooling technology and the highest capacity for heat rejection in a given footprint. When you can cool high performance servers in such little space, the edge can proliferate across the world, enabling the technologies of the future. Space may be limited, but potential is not. 

Stay in the know!

Sign up to get notified about the latest LiquidStack news, resources and events.