DriveScale’s founders wanted to solve the problem of infrastructure over-provisioning when they started their company. But the technology wasn’t called composable infrastructure back then.

“That was before 2013, and they were looking at the changing data center applications and workloads being created in the hyperscale clouds,” said Denise Shiffman, chief product officer at DriveScale. These were workloads like Hadoop, Casandra, and Spark, and they behaved differently than traditional data center workloads like Oracle and SAP. “These applications were much more dynamic and the demands for latency were very high, while the history had been running predictable applications in the data center,” she said.

In addition to performance problems, data-intensive workloads also lead to server sprawl. Companies can’t predict how much storage and compute capacity they need. So they over-buy. “Many customers we talk to are really only using about 20 percent of the infrastructure they’ve deployed,” Shiffman said. “The [DriveScale] founders believed there was a new way to fix this and that was taking storage outside the server and putting it on the network.”

Duane Northcutt, Satya Nishtala, and Tom Lyon founded DriveScale in March 2013. Northcutt, VP of engineering at DriveScale, previously worked as chief technology officer at other tech companies including the Connected Home Division of Technicolor, Trident Microsystems, and Silicon Image. Nishtala, DriveScale’s chief architect, is a former Cisco distinguished engineer. And Chief Scientist Lyon previously co-founded several companies including Ipsilon Networks and Nuova Systems, which was acquired by Cisco and became the basis for its UCS servers.

They raised about $15 million in a 2016 Series A funding round.

About 18 months ago the startup launched the DriveScale Composable Platform. The software-only product lets users create their own server platform on-the-fly with the exact compute and storage resources needed for a given application or workload. It’s hardware agnostic and can use different vendors’ compute, storage, and network fabric.

The platform disaggregates compute and storage from the traditional direct-attached storage server and connects to a standard Ethernet network. It supports bare metal or containerized applications.

The product targets six use cases: big data analytics; machine learning; tiered data storage; cold data and object storage; all-flash; and containers and Kubernetes. “We think applications are going to move toward Kubernetes and we think that’s a huge opportunity,” Shiffman said.

The platform integrates with both Kubernetes and Docker, providing persistent storage for container environments. “I believe we are the only composable infrastructure vendor that has a Kubernetes plugin," Shiffman said. The popular container orchestrator is ideal for unpredictable applications and workloads and is designed to be dynamic by nature, she explained. “It’s designed to optimize the infrastructure. So together Kubernetes and DriveScale provide highly optimized utilization of the resources underneath.”

DriveScale started shipping its product about the same time in 2017 that Hewlett Packard Enterprise (HPE) made available Synergy, its composable infrastructure. HPE was the first traditional infrastructure vendor to bring composable to market — others including Dell EMC and Juniper Networks have since added composable infrastructure to their portfolios. And HPE has said it’s the company’s fastest-growing new technology category with more than 1,600 customers.

During its fourth-quarter 2018 earnings call this month, HPE CEO Antonio Neri touted Synergy’s “incredible year,” growing more than 280 percent and reaching an annual run rate of over $1 billion.

While DriveScale competes with HPE in this space, Shiffman said the startup’s software form factor gives it a competitive advantage. “HPE did a good job of defining Synergy, but it is tied to their hardware — and expensive hardware at that,” she said. “One of the core capabilities you want with composable infrastructure: it should make low-cost, commodity compute and storage highly available, high performance, and very high scale. We’re truly a software application that runs on anyone’s hardware.”

It can scale to more than 1,000 compute nodes and over 100,000 drives, she added. Also, DriveScale built dual ports, load balancing, and encryption into the platform. “All of those things differentiate us from pretty much anyone else trying to do this.”

The company has a “handful” of customers, and some are large enterprises, Shiffman said. This includes digital advertising company AppNexus, which AT&T acquired in August, and health care analytics company Clearsense. “We have trials going on all over the world: in China, Japan, Europe, and the U.S.” Shiffman added.

It also has significant partnerships with Cisco, Dell EMC, RedHat, and Cloudera and Hortonworks, which recently merged, and other technology companies.

As enterprises move further into data analytics and machine learning and want to make better use of the massive amounts of data they accumulate, Shiffman expects to see more demand for DriveScale’s software. “As companies move to using their data and getting more info out of the data than they historically have, composable memory and compute and other technologies really makes sense.”