The Problem

As a cutting-edge, deep learning analytics company, Vyasa's demands for computing power are much higher than the average organization. However, with greater computing power comes higher costs. As a startup, the company knew it had to get up and running as quickly as possible, so when it came time to make a strategic decision on its IT architecture, it considered two options—a cloud-only approach or a hybrid architecture.

Many organizations embrace the cloud for flexibility and scalability, but for Vyasa, that would mean running all of its high-compute R&D activities on machines it didn't own, producing an unpredictable monthly stream of costly bills. Alternatively if it went with a hybrid architecture, the company would be required to make investments in its own hardware, as well as to find a location to store, run and manage the systems infrastructure.

According to Christopher Bouton, the CEO of Vyasa, the right decision quickly became clear. "As a data science company, it was going to be critical for us to employ a hybrid architecture of our deep learning infrastructure. In particular, for all sorts of R&D purposes, running all of our workloads in the public cloud would get prohibitively expensive."

The Solution

After an introduction from his CIO, who previously worked with Markley, Bouton quickly knew that the Markley One Summer Street facility—New England's largest and longest operating mission-critical data center and telecom hub—was what Vyasa needed.

Vyasa knew they needed an advanced GPU infrastructure and the ability to continually research and test new systems and algorithms. With Markley, Vyasa has the secure, reliable lab infrastructure they need to innovate and advance what's possible. "Doing that on our own systems at Markley versus in the cloud provides us with much greater control over our resources and the ability to iterate on our algorithms and data sets more quickly without incurring more expense," Bouton said.

Today, the Vyasa development lab is housed in the Markley data center with 2N uninterruptible power and cooling, with high-performance Direct Internet Access (DIA) connectivity back to its corporate headquarters. The company has also benefited from their direct relationship with Markley, which provides access to Markley's technical teams for support and expertise across high-performance computing, networking services, expansion of racks, additions of new hardware and more. All of Markley's technical capabilities are important, but aren't the only things Vyasa considers valuable.

"One of the key things Markley did for us, perhaps surprisingly, was to make us feel like family. There's this incredibly welcoming spirit in the Markley data center," said Bouton. "Not only have we been able to utilize the facilities from a technology standpoint, but Markley welcomes us to hold meetings there and connect with our clients. Based on our initial interaction which was so positive, it was an easy decision to go with Markley."

The Results

With its lab housed at Markley, Bouton is confident that Vyasa has been able to reduce its compute costs to levels they'd never otherwise see by committing to a cloud-only approach. Vyasa can continuously run R&D projects on its in-house systems for what quickly has become a more economical solution, even on relatively expensive machines.

For example, with help from their inclusion in the Mass-Development Emerging Technology Fund, Vyasa recently purchased an NVIDIA® DGX-1™, a six-figure, high-performance, deep learning server. By hosting it at Markley, Vyasa has fully recouped its hardware investment in less than a year, as compared to running a similar level of compute infrastructure in the public cloud.

Conclusion

With its infrastructure at Markley, Vyasa has been able to achieve the hybrid cloud infrastructure needed to power its deep learning algorithms and R&D activities—at the price point the company demands. Beyond that, Vyasa has found a dedicated partner who cares about the company's goals and mission.