Falkonry provides AI services to its industrial customers generating great volumes of IoT data. Here's why this company chose a bare metal cloud implementation for its operations.
6 Signs You Need A Multi-Cloud Strategy
(Click image for larger view and slideshow.)
Getting fast answers is a top job for information-intensive businesses that ingest, process, and analyze vast volumes of data. While adding more data and more complex algorithms can reveal richer insights, it can also slow down how fast the answers can be delivered.
Artificial intelligence service provider Falkonry grapples with this challenge every day. It works with IoT data from industrial companies, and looks to provide insights that can help its clients get value out of those vast streams of proprietary data. The results might include improved quality control, greater efficiency, and the ability to predict problems before they happen.
For instance, Falkonry can help a chemical manufacturing company identify ways to improve quality control before the output can even be tested for quality. The manufacturer can then leverage that knowledge to make tweaks to its process sooner.
The high volume of data and the need to get deep insights at a rapid pace are what has driven Falkonry to choose bare metal implementations of its stack.
"AI can be accelerated when we have some proximity to the hardware," CEO Nikunj Mehta told InformationWeek in an interview. "We believe that, for our continuing success, being close to bare metal is important."
Mehta said he believes that any IT organization that is pursuing an AI project to help its business achieve its goals would benefit from bare metal implementations. These implementations can provide benefits to IT organizations, too, Mehta said.
"Cloud providers have become encampments for customers who are unable to move from one cloud provider to another," Mehta told InformationWeek. But these workloads are "more portable and more economical when deployed on bare metal."
Ovum chief research officer Tim Jennings puts it this way: "In some cases, enterprises simply want to 'lift and shift' a workload to the cloud with minimal change," he wrote in an email response to InformationWeek questions. "Bare metal IaaS largely gives them the option to do that -- i.e. the customer makes the choice of operating system, middleware, database, etc., right up the stack."
Today, Falkonry operates its bare metal AI in the cloud. The company had been working with a technology stack that leveraged open source technologies including Apache Spark and Google's open source container orchestration technology, Kubernetes. Earlier this year it began to investigate options for bare metal cloud providers.
At the beginning of this year, Oracle's Bare Metal Cloud emerged as the strongest candidate to host Falkonry's AI, according to Mehta. Oracle took the wraps off this its bare metal cloud offering during its Oracle OpenWorld event in September.
Gartner VP and distinguished analyst Lydia Leong said in a blog post that Oracle's bare metal cloud "is a true software-defined cloud IaaS offering, provisioned in minutes and billed by the hour. This sets it apart from more hosting-like bare-metal offerings such as IBM SoftLayer, Internap, and Cogeco Peer 1."
The way Oracle structured its bare metal cloud offering was a fit for Falkonry, Mehta told InformationWeek.
Mehta said that the Oracle cloud worked well for its requirements of open standards and support for the networking requirements of hybrid clouds. The transition to Oracle Bare Metal Cloud five months ago was a smooth one. Since then, Falkonry has been able to increase the amount of work it's running for customers.
"Falkonry has had to design its deployment to work in a number of different environments," Mehta said. "We didn't have to perform any changes to our architecture to move to Oracle Bare Metal Cloud."
Is bare metal cloud generally a good choice for data-intensive workloads such as AI and machine learning? Not necessarily, according to Ovum's Jennings.
"In general, I would say that bare-metal cloud is not better suited for those tasks," Jennings told InformationWeek. "Areas like machine learning, AI, and analytics benefit from having a range of relevant platform services on top of the bare infrastructure with which enterprises can build their applications."
Jennings did see a counterargument, too.
"In some of these applications, the need for raw compute power (particularly in-memory processing) means that, for extreme performance, some organizations will want maximum control over the infrastructure and ability to tune it," he told InformationWeek. "In that circumstance, bare metal can be the preferred option."
Jessica Davis has spent a career covering the intersection of business and technology at titles including IDG's Infoworld, Ziff Davis Enterprise's eWeek and Channel Insider, and Penton Technology's MSPmentor. She's passionate about the practical use of business intelligence, ... View Full Bio
We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.