Demand for computing power will continue to grow as AI and other technologies evolve. Canadian startup Distributive has developed a new distributed computing technology that uses idle computers, enterprise IT infrastructure, and Internet of things (IoT) devices. The solution can help reduce cloud computing costs by up to 92%.
Dan Desjardins, Distributive's CEO, said the concept of distributed computing is splitting a task into many small pieces and distributing them to multiple computers, servers and IoT devices for simultaneous execution.
Desjardins said the technology is currently enabling hospitals, universities, airports and other organizations to tap into idle computing power from the computers they already own to work on big problems like AI, big data, and advanced research computing.
"Rather than bringing the data to the compute, which is the cloud model, we bring compute to the data," Desjardins said.
A solution that slashes cloud computing costs
Distributive has developed the Distributive Compute Protocol (DCP), a compute platform built on web technologies. According to the company's website, the platform can match the supply and demand for compute to access abundant computing resources.
Desjardins said with the explosion of AI and digital transformation, demand for computing resources is beginning to outstrip the supply available from the cloud. The situation also puts pressure on chip manufacturers to increase production. Meanwhile, millions of computers worldwide are idle for at least 16 hours of the day and can be tapped into for compute.
DCP enables distributed computing on any server, computer, or IoT device, regardless of operating system or hardware, according to Desjardins. The company has run the solution on smartphones, smart watches, VR headsets, Tesla cars, laptops and many other devices, demonstrating maximum interoperability.
One of the most significant benefits of Distributive's solution is that it can trim the cost of cloud computing by 92%. Desjardins said the company delivered a computer vision solution on top of its computing platform to an airport in Canada. The airport wanted an automated ground safety monitoring and alerting system.
Desjardins noted that commercial cloud machine vision solutions like AWS Rekognition or Microsoft Custom Vision charge approximately US$100 to perform 20,000 image inferences.
"Assume that your camera only generates one frame per second, which is low. The cost of continuously running Rekognition or Custom Vision is (close to) US$160,000 per year," he added.
Desjardins said Distributive's distributed machine vision platform, Overwatch, performs the same function for only US$12,000 a year, a 92% reduction compared with the commercial solutions.
The more affordable technology will be able to help companies mitigate the burdens of cloud bills, according to Desjardins. The industry estimated that about US$800 billion will be spent on computing power in the cloud by 2028, largely due to exploding demand for AI products.
Desjardins said Distributive's solution also interoperates with the cloud as needed. At the same time, it intends to be an extra option to compete with the cloud for a better price, simpler use and data privacy.
An application that increases hospital surgical efficiency
In addition to its distributed computing platform, Distributive has begun commercializing a solution called Osler for hospitals. The solution is a surgical block schedule optimization application, according to Desjardins.
He explained that Canadian hospital administrative staff are responsible for generating block schedules for several months at a time. These block schedules lay out what surgical services occur in which operating room on which days.
According to Desjardins, depending on the number of operating rooms and surgical offices to coordinate with, planning a six-month surgical block schedule encompasses thousands of decision points. It can take up to three months in some cases. To streamline the process, Distributive created an application running on its platform that computes all possible block combinations, predicts their performance and selects the most efficient schedule, Desjardins said.
He added that Osler minimizes resource contention and increases efficiency by as much as 10%. The process only took seconds instead of months. Desjardins said the solution has been deployed in five hospitals across Canada.
"What's really cool is that it uses computing power from computers in the hospitals, and none of the data leaves the building," he said.
The CEO also stated that Distributive offers its distributed computing tools and software development kits to everyone. The goal is to allow compute-intensive applications to be built and operationalized for a fraction of what would cost on commercial cloud computing platforms. The company currently provides the platform at no cost to academic institutions worldwide.
Desjardins said that Distributive's mantra is "give to academia, commercialize with industry."
He also said the company is engaging with institutions in Brazil, Kenya, the UK and Ireland. Now Distributive has come to the Asia Pacific region, with Taiwan being the first stop. Desjardins said as Taipei holds remarkable technology and ICT capability, it is a natural fit for the company to explore the market and see how the innovation is received there.
