Larry Ellison made a splash this week when he said that Oracle would give Amazon a run for its money in the cloud. Then, the company outlined the pillar of the tech titan’s infrastructure offering: beefy, bare-metal servers running in the cloud.
That’s right: Oracle is going after a market that’s full of virtualized workloads with servers that clock in with a whopping 36 physical CPU cores, according to Vice President of Engineering Don Johnson. Rather than starting from the low end of the infrastructure market and working its way up, Oracle is starting at the top and working its way down.
It’s a move that’s aimed at power-hungry enterprise workloads that require a lot of processing power, like Oracle Database or video rendering. The company will provision bare-metal servers when customers ask for them, and customers are billed by the hour for what they use.
The strategy has earned some early praise. “Smart and scalable choices seem to have been made throughout,” Gartner Vice President Lydia Leong said in a note. “However, I would characterize this early offering as [a] minimum viable product; it is the foundation of a future competitive offering, rather than a competitive offering today.”
Oracle is making a play that takes advantage of a lot of its strengths in the cloud market. The company has access to the latest and greatest hardware, and it builds all the hardware it uses, said Deepak Patil, a vice president of product development at Oracle.
Oracle’s infrastructure moves are well-suited for the company’s host of on-premises customers, many of which haven’t migrated their workloads to the cloud yet, said IDC analyst Al Hilwa.
“They have especially not moved their large enterprise databases, which typically run on bare metal, and indeed, large VMs are likely needed to run the typical Oracle database,” he said. “Oracle’s focus on the high-end is a good fit for its customer-base but is predicated on the fact that the customers are willing and able to move.”
One of the key architecture advantages for Oracle’s bare-metal offering is the fact that the company is running network virtualization on the network layer, and not in the hypervisor. That allows the company to hand users the keys to a completely bare-metal server that Oracle isn’t running software on, in addition to other network connectivity advantages.
Customers will be able to set up whatever operating system and hypervisor they want on top of the hardware, Johnson said Monday. Some configurations will require more work than others, but the guiding principle is that the servers will indeed be bare metal.
Having a bare-metal server also provides security benefits, Johnson said. Because Oracle handles all of the network virtualization work away from the bare-metal server, users have physical isolation of workloads from other cloud tenants and the provider itself, a feature exclusive to this cloud offering, he said.
“Not only do you not need to trust or worry about adjacent contents, you don’t need to worry about the provider,” he said. “If there’s zero trust, there’s less risk.”
One of the key issues that Oracle faces with its infrastructure rollout is scaling it up worldwide. Right now, the infrastructure offerings are live through the company’s western U.S. region, based in Phoenix, Arizona. The company is aiming to have its eastern U.S. region running by December or January and expects to expand to two European regions by mid-2017.
The company seems committed to the infrastructure push, IDC Group Vice President Al Gillen said in an email. But scale will be an issue.
“It is also a long-term commitment that won’t be ready for anything beyond test deployments until a second zone becomes available in the U.S. market,” he said.
Gartner’s Leong had a different take, saying that scale isn’t as much of an issue for the company as it is a problem with its quality of offering and a current lack of customers.
“Oracle needs to build a compelling feature set, get customers, and then expand data center footprint,” she said.
Oracle seems committed to the hard work of building out its offering, but executives at the company have different takes about the shape that takes. When asked Monday about the company’s infrastructure spending, CEO Mark Hurd led with the company’s SaaS business.
“So, if we had a choice strategically, we wanted to nail the SaaS business,” Hurd said. “Then the second thing that we wanted to make sure that we got right was the PaaS business. You’ll see more investments into infrastructure, but remember, it’s riding behind the wake of the SaaS and PaaS infrastructure that’s out there.”
Johnson, meanwhile, led with the company’s infrastructure commitment.
“The key foundation of a cloud platform is infrastructure. It’s IaaS,” he said. “It is the fundamental building block that makes everything else possible. [With] all the things that you require up the stack, either the foundational pieces give them to you, or you don’t have them.”
The two perspectives aren’t incompatible; after all, those SaaS applications have to run on something. But differing views on Oracle’s business have appeared: one a SaaS- and PaaS-first cloud that’s building an infrastructure business, and another an infrastructure company using that expertise to operate SaaS and PaaS products.
It will be worth watching Oracle to see how that tension resolves and how cloud customers and competing providers react to the company’s plans. Competitor Amazon is expected to release new high-performance compute instances later this year.