Could smartphones replace datacenters? These Finnish researchers think so

Researchers think they've found a way of using the vast untapped resources of processing power in smartphones and IoT devices for large-scale number-crunching.

phones.jpg

The Ubispark researchers tested their software using Samsung Galaxy S4 phones and a LG Smart TV and found them more than the equal of a single server in computing power.

Image: Ubispark

A smartphone today packs more computing power than the computers used by NASA in the Apollo space program. At the same time, the Internet of Things is bringing connectivity to a growing number of devices, from smart TVs to fridges.

Now researchers at Finland's University of Helsinki want to harness all these computational resources in a project called Ubispark, which uses smart devices for energy-efficient distributed computing.

Smartphones in 2017: What's in the pipeline?

The latest rumors and trends in the smartphone industry. Apple, Samsung, and a few smaller players like OnePlus have a few tricks up their sleeves.

Read More

The aim of Ubispark is to create local computing clusters made up of phones and other smart devices that collectively can run large-scale data-processing tasks.

"You have connectivity through mobile networks, which is available more widely than the fixed networks you need for cloud computing or to have your own datacenter," Eemil Lagerspetz, head of the research project, tells ZDNet.

"The smartphones in our pockets are even faster than [the supercomputers of the past]. There is the power, there is the connectivity, so why not [use them for computing tasks]."

The goal of the research project is to build a mobile platform that enables smartphones and IoT devices to run computing tasks that traditionally require larger computers and datacenters.

A user downloads the app onto their device, chooses which scientific or commercial projects they will allow their phone to be used for, as well as the conditions, such as only running when connected to Wi-Fi or a charger. One hour of computation consumes an estimated 10 percent of battery life.

"The app would then connect to the participating projects that want to offload computing to smartphones," Lagerspetz explains. "Those tasks would start coming in from the cloud into the phone and run there and the results would be sent back to the originator of the task."

In other words, while your phone sits idle on your office desk, it could be helping to compute weather forecasts, earthquake warnings, or solve encryption challenges.

But first the mobile cloud computing platform needs to be built. Currently the researchers are testing their approach using nine Samsung Galaxy S4 phones and a LG Smart TV for predefined tasks. The preliminary results show that this configuration is more than enough to equal the computing power of a single server.

"We have nine smartphones computing in parallel and one server computing the same thing, and we can achieve the same speed," Lagerspetz says.

"[But] if you want to beat a cluster of machines at Amazon, you might need 10 times the number of phones, so 90 phones. If you're working in a big organization and you have a lot of employees, you just put our app on their phones and you're ready."

This approach is not only a cost-effective alternative to running your own datacenter or paying for cloud computing, but also saves energy. According to Lagerspetz, smartphones are typically 50 percent to 90 percent more energy-efficient than a server.

While Lagerspetz aims to create a flexible mobile platform for cloud computing, the idea of crowdsourced processing power is not new.

For example, Seti@Home and IBM's World Community Grid projects use the Berkeley Open Infrastructure for Network Computing, or BOINC, platform to harness the unused computing power of computers, smartphones and tablets for scientific research.

"BOINC has a star-like architecture so there is no connection between the different devices, they all have their own connection to the central server," Lagerspetz explains.

"The idea in our project is that devices in close proximity can share intermediate results and run multi-stage computing tasks in cooperation, independently from any server."

Lagerspetz says using phones or IoT devices in the same location, such as an office space, is optimal for achieving the best computing results. The devices can have local fast connectivity between them and act as one unit.

As the distance between the devices used for a computing task grows, the more problems arise as data transfers and the varied quality of mobile networks start to become more of an issue.

It can also be a challenge to convince people to give access to their devices. Lagerspetz hopes to create a similar win-win scenario to the smartphone battery research project Carat, where users get details on what their battery problems are in exchange for sharing data. Alternatively, the incentives could be financial, if the open-source platform is used commercially.

"For example, electricity companies or mobile operators could give you some benefit from giving your computing power to them while your phone is not used," Lagerspetz says. "They could pay a part of your electricity bill or give a discount on your monthly payments."

While there are still several challenges to tackle, including the limitations of current mobile platforms, Lagerspetz hopes to have the first version of the app available for Android this year.

He believes that the untapped power of smartphones and IoT devices has a huge potential in taking computing to where people are and eventually even replacing datacenters.

While we are asleep, it could be our phones, TVs, and washing machines that are at work helping scientific research.

cluster-laptop-server-execution-time.jpg

The Ubispark researchers found that the computation time for the same task for a cluster of smartphones is higher than that of the server, for just 7.2 percent of the power.

Image: Ubispark

Read more about technology in Scandinavia