The classic datacenter model was based on Moore's law. But this only promotes the status quo, and there are three primary shortcomings to the model. Read what they are and why hyperscale provides the best way forward.

IT infrastructure has long evolved using the same  classic model, in which existing floor space, wiring, power and cooling has been leveraged by more and more powerful server blades to deal with capacity growth. This model worked well in an age when IT infrastructure was seen as a cost center without strategic business impact.

This is no longer true.

The classic datacenter infrastructure model

The classic infrastructure model is anchored in Moore’s law, with an expectation that technology evolution will be fast enough to match capacity growth demands now and in the future.

IT infrastructure was then deployed in centralized datacenters. Once you deployed cabinets, wiring, power and cooling, you had the basics in place. The rest of your journey was just about introducing new blade servers at the pace capacity needs grow.

This classic model was simple to plan for, as blade prices have been stable for the last decade and their power consumption has not changed. Server blades are depreciated over 2-3 years while the rest is depreciated over significantly longer periods. And last but not least, IT workloads have grown with predictable growth rates.

The shortcomings of the classic datacenter model

The classic model tends to conserve a status quo in your datacenter and this is putting your future digital competitiveness at risk.

There is no/low incentive to minimize the resources used, as most resources are viewed as fixed. You conserve the historical footprint, power and cooling frame you deployed in the previous decade. Leveraging fully depreciated resources is a sound strategy in the classic model, even if itcan drive sub-optimizations.  It’s not a sound strategy any longer.

The second primary shortcoming is the actual use of IT resources. Resource utilization rates are often hampered by fixed relations between compute, memory, storage and networking resources. This is a direct result of classic solutions designed to always be backwards compatible with the cabinets, wiring, power and cooling put into your datacenter in the first place. 

Shortcoming number three is about the human resources required to operate the datacenter infrastructure.  With all parameters stable, it is hard to achieve any efficiency gains

The classic model worked once upon a time but now it has severe limitations, as workloads are becoming more diverse and infrastructure is becoming the foundation for your digital transformation. This represents a conundrum crucial for your future infrastructure agility – a conundrum in which you should question whether Moore’s law should be your only recipe.

Hyperscale datacenters – the alternative route forward

The alternative is to approach your infrastructure from what you want in the future, rather than from what you had to the past.

Take a look at a software-defined infrastructure model, starting with a detailed understanding of exactly how your current infrastructure resources are used. This gives you a holistic view on IT, human, building, wiring, power and cooling resources, as well as a strategy for resource optimization across all of them.

A transition to optical wires in your datacenter

Moore’s law will drive technology development to a point where your existing wires get outdated. You can still use classic wires but the instances in your datacenter will get smaller and smaller and work against your resource utilization targets. Fiber technology advancements have now reached a point where they represent the current and future preference.

Datacenter locations will move in two different directions:

  • larger, more centralized and cost optimized for non-real time applications
  • smaller, more distributed and performance optimized for real-time and latency critical applications.

Digital transformation will affect all operators and all enterprises going forward. It is a challenge that requires new perspectives and approaches for how we build infrastructure.

The time has come to for you to evolve from being a Rack-star into the hyperscale hero your business needs to get the future right.

To explore these ideas further, read our white paper on Hyperscale Cloud - reimagining datacenters from hardware to applications:

Download the white paper


Cloud Infrastructure Digital Industrialization

Peter Linder

Peter Linder is Head of Business Management and Sales Support for Business Unit IT & Cloud Products towards Region North America. Since 2011 Peter has been based in North America in various management roles for the development of Ericsson’s cloud and IP Business in the US and Canada. He is also a Network Society evangelist appointed in the original group in 2011 and an intrapreneur dedicated to learning and sharing insights on how the digital transformation is reshaping future networks.

Follow Peter Linder:
Peter Linder

Discussions