Sergey Nivens - Fotolia

Nano server innovation extends beyond the data center

What's the difference between a nano server and a PC? Timing.

Data center servers depreciate over several years until they hit the e-waste pile, but there's a new concept for compute that can have multiple incarnations.

Nano servers -- not to be confused with micro servers -- combine industrialized, modular computing components for use as data center servers. Then they can be used as high-performance desktops or tablets, and finish their days as PCs. This means the data center's computing infrastructure refreshes at a rapid pace and combines the purchasing power of end user computing and data center IT teams.

Jacob Hall, innovation wrangler for Wells Fargo, a U.S.-based multinational banking and financial services provider, leads and manages the company's internal technology incubator, Wholesale Labs. He's explored the benefits of nano servers and spoke with about facility operations and performance payoffs of this innovation, server lifecycles and more.

Why nano servers?

Jacob Hall: Organizations buy hundreds of thousands of desktops every year. Many companies refresh end-user devices each year. This number is growing even more for tablets.

Computers get faster and more powerful quickly. Nano servers combine modular blocks of core computing architecture. With this idea, IT shops could buy the newest computers and put them in the place with the best energy efficiency, where computers run 24/7: the data center. They can then take these components from the server out of the data center and use as desktop, a tablet, smart display, wearable computer, et cetera.

Nano servers modularize the "guts" and combine them into [these other form factors] over time.

In what ways are nano servers more efficient than using x86 servers? Aren't tablets and mobile devices usually based on RISC ARM devices and servers on CICS x86 processors?

Hall: It's far easier to maintain the same computing architecture -- all x86 or all-ARM -- that an enterprise already uses. The architectural change of modularizing the server units is abstracted from processing architecture.

Jacob HallJacob Hall

You wouldn't try to replace everything in the data center with nano servers. There's still a place for big iron for heavier workloads. But for virtual desktops, Web servers, high-performance computing environments [and so on], these devices make a lot of sense. A database server can run on a laptop, which means nano servers can even run small database servers, or you can spread data over many nano servers.

Is that enough computing power? Consider how virtualization carves a processor up into many virtual servers. Virtualization fractionalizes compute resources, and it needs all this hypervisor software to manage the process. Nano servers do the reverse, putting smaller amounts of processing power together to support workloads without having to worry about workload sharing.

Is it more cost effective or does cost not factor in to the equation?

Hall: This is a method to run a more energy-efficient data center, which means better costs. At better performance per watt, [the end] customer experience is better. Essentially, you're transferring the spend on energy and cooling to better performance-per-watt technology. All tech companies should want this to happen.

If I had a blank check and someone asked: 'How would you revolutionize computing?', this is what I'd do.

I call this cold computing, because if you add up the total amount of heat created by traditional compute designs and compare that to the total heat created by a modular compute design, the modular design is colder overall. With modular computing, the workloads that demand the most [resources and fastest response times] rapidly receive better hardware upgrades, rather than waiting years for [them].

Typically, companies will buy a server and depreciate it -- leave the hardware running in the data center for a long time, because it can't be repurposed for another use -- so they run less-efficient devices that output a lot of waste heat and have a higher energy cost to operate.

With [the nano server] concept, you transfer what you would have spent on cooling costs into higher performance sooner. ... A modular compute design that works across many form factors increases the feasibility that compute devices are repurposed out of data centers faster. Waterfalling off devices will also give a wider reach of people access to the Internet faster -- getting the next billion users online sooner. The processor [could go] from server to desktop to home device to personal device in emerging countries to school computer.

Where do you see traditional IT hardware vendors in this new design?

Hall: Industrialized computing is the next step of commodity computing -- an array of slots, plug in a new piece when an old piece breaks.

This could help turn around the hardware vendors' roadmaps -- an inflection point that changes where money is allocated and what people buy.

It is already happening to an extent with microservers. So what's wrong with the microserver design? You can't repurpose them. The rapid investment doesn't continue to pay back.

Why go with consumer-grade over enterprise-grade hardware?

Hall: There are other ways to handle reliability in enterprise data centers than hardware. People run servers without ECC RAM and it works just fine. It is possible to build reliability into the software, and not make everything one-to-one redundant. My [operating system] can move to another machine if the hardware fails, for example. You used to spend a lot of money on a highly reliable server with mirrored RAM and ECC -- for most use cases that isn't necessary, particularly with modern software design. This is a direct way to reduce cost.

How do you compare this data center vision to cloud?

Hall: From a corporate perspective, there are challenges with moving to the cloud. Nano servers are the answer to maintaining your own [data center] infrastructure and performing closer the level of efficiency of public cloud.

There aren't many ways to become more efficient, but buying commodity compute at rapid refresh cycles and reducing energy/cooling cost is one. By combining two purchasing teams -- end-user devices and data center -- an IT organization can get more volume from the same IT vendor, improving prices.

If I had a blank check and someone asked: 'How would you revolutionize computing?', this is what I'd do.

Next Steps

Discover data center innovations from around the world

Data center architects share their latest innovations

Dig Deeper on Server hardware strategy