The data center is a crucial component of the global digital ecosystem that supports the hardware, software, and networking systems of various Internet services and business applications. With the indefinite yearly growth of the requirements to process, store, and access data, data center racks must also advance.
Rack servers are becoming more compact and are also increasing their capabilities and density with increasing rack density and new designs and architectures that appear on the scene to fill the need for more capacity.
Let’s explore the top 8 rack technology trends that are shaping the growth trajectory of future data centers.
1. Key Drivers for the Hyperscale Data Center Design Accelerate
Hyperscale, also known as hyper-scale, data centers are large-scale buildings constructed by companies such as Amazon, Google, Facebook, and Microsoft to meet the extremely growing demands for computing services by these companies.
As for the architectural strategy of hyperscale data centers, it applies certain measures of scale to achieve optimal density and the greatest concentration of calculating power over the least area.
Key hyperscale design trends include:
- Modular pre-fabricated elements of construction are used in rapid growth of facilities.
- Less focus on air recirculation and “free cooling”.
- Removal of raised floors to accommodate more racks of equipment.
- Advanced power railway, micro-grid and smart distribution systems.
- Delayered networks and optimised rack server layouts.
2. Rack Servers Evolved into High-Density Designs for Data Processing Needs
Before the advent of the blade servers, the racks of the data centers used to accommodate 1U or 2U servers at best. High-density, low-priced rack servers are now packing more and more computing into a small housing package. Today’s blade servers or modular ones can fit up to 20U or more in one rack space!
Density-optimizing server designs include:
- Half-wide server racks.
- Multi-node servers.
- Blade enclosure racks that are cooled in the vertical plane.
- High-powered racks are usually placed in immersion cooling tanks.
Such density improvements increase computing density per square foot; these benefits cut data center expansion expenses. Cost savings on energy are also realized from the sharing of resources across heavily clustered hardware.
3. Mainstream Adoption of Liquid Cooling
Historic data center cooling employs perimeter CRAC units and a raised floor to blast cold air to server intakes, cooling essential equipment. However, these kinds of ‘air cooling’ techniques have proven to be inadequate to accommodate the amount of heat generated by processors today.
Direct-to-chip liquid cooling systems use pipes for circulating water or coolant directly over chips and components through heat exchangers located in server racks.
Today’s racks for homes have emblematic rear-door heat exchanger layouts that do not incorporate IT equipment. Fire suppression systems have equally advanced in a way that they can safely handle leakages.
4. Sustaining of Innovation in Power Distribution in Rack-Based System
Higher rack power densities generate greater demands for energy distribution and subsequent service, making it critical to prevent expensive outages. Modern rack-based power systems offer precision control and detailed monitoring, as well as remote management and automated policy for near 100% availability and effectiveness.
Key power innovations shaping future data centers include:
- Increased distribution of power from the rack level by using 400 V or 480 V supplies.
- Lithium-ion battery backup units are installed in rows.
- 10G rack transfer switches for concurrent power feed redundancy.
- Combination of IT, power and cooling equipment in a single rack.
This sort of rack power improvement, for instance, offers the high power density and reliability required in compact, high-availability situations.
5. Disaggregated Server Architectures: The Key to Growth
Conventional rack servers integrate processors, memories, flash storage, and accelerators into single housings. Conventional architectures integrated those tightly coupled resources into one system, while disaggregated computing architectures divide them into multiple pools that may be dynamically combined depending on the rack in data centers.
Disaggregation promises benefits like:
- Independent hardware refresh cycles are those that are not linked to an OS upgrade but are instead carried out independently in order to refresh the physical layer of the hardware.
- Constant resource mobilization and redistribution.
- Incremental, pay-as-you-grow scaling.
In a similar fashion, vendors are advancing sliced server and rack architectures with interconnects like PCIe to offer composable resources to rack pools. If more extensive use of disaggregated racks is achieved, the data center will offer the same flexibility as cloud services and more efficiency.
6. Cloud-Native Architectures Continue to Gain Momentum
The newer cloud-native technologies of containers, microservices, and Kubernetes for container orchestration are increasingly permeating beyond the cloud and into on-premises data centers.
Cloud-native platforms provide:
- Being able to run specific applications across different environments.
- Immutable infrastructure patterns.
- Autoscaling based on demand.
- Infrastructure-as-code automation.
Opportunities like these cloud-aligned models are making changes to the rack server for home designs, the storage, and the data center networks. Racks are becoming converged, hyperconverged, and composable to act in a cloud-native fashion. Layouts, power configurations and how cooling is conducted are also driven by the transition to cloud-native data centers.
7. New Accelerators for Artificial Intelligence
AI workloads involve the necessity of perfect synergy with the match between the application and the hardware. End computing includes GPU, FPGA, AI accelerator cards, and general-purpose server CPUs within the same rack.
The increase in parallel compute industries that include AI, ML, and analytics will increase demand for heterogeneous racks that align applications to the right processing platform. Today, solid-state vendors provide heterogeneous blocks containing multiple CPU, GPU, FPGA, and cache abilities in 2U packages.
8. Actively Implement Standardized Rack Management Frameworks
Hyperscale operators have initiated efforts in open rack server management frameworks like Open19 and SNIC (Scalable Negotiation Infrastructure for Cloud) to make the racks standardized for efficiency. These open standards prescribe the areas for rack height and thickness, power consumption, and networking interface of shared racks.
Open-rack standards promise benefits like:
- Relaxed vendor-specific hardware procurement policies.
- Interoperable multi-vendor configurations.
- Enhanced asset protection and optimised lifecycle management.
- At the rack level, the failure domain is isolated.
A standardized rack layout will also ensure that new technologies can be incorporated into a new generation of data center architecture.
Final Words
The data center rack is at the crossroads of vital data center infrastructure, including power, cooling and networks. Fueling those auxiliary systems’ rapid advancements are increases in rack density, which delivers exponentially higher compute power per square foot.
Designs inspired by hyperscale facilities, combined with intensive cooling methods and the separation of rack elements, represent trends that will determine further rack developments in the following 5–10 years. It will complement the innovations of cloud-native platforms, AI accelerators, and open rack standards over an even more optimized but scalable data center infrastructure.
Read More: The 12 Most Important Things to Know About Rack Servers
