The recent GigaOm Structure 2014 conference in San Francisco drew more than 700 participants for two days focused on cloud security, the Internet of Things (IoT), new architectures, and DevOps. Participants from GE Software brought back the following key takeaways:

Cloud Computing

The Target debacle was a wake-up call for all retailers,” commented Jeffery Padgett, senior director, infrastructure architecture at Gap, Inc. The good news, per Florida Crystal VP and CIO Don Whittington, is that Cloud security can be much better addressed today than four years ago. And Amazon’s CTO Werner Vogels seemed to agree, pointing out that customers could (and probably should) encrypt all the data they store in the Cloud.

Vogels also believes that stringent regulatory requirements (such as Europe’s) can be addressed effectively by US providers using local cloud storage options and by following “good architecture principles,” supported by clear deployment and operating blueprints. Per Vogels, those security efforts are paying off: The CIA is an adopter of Amazon’s Cloud, and so is GE, running some of its Predictivity applications in the Cloud, deployed on CloudFoundry PaaS [underpinning Predix, GE’s software platform for the Industrial Internet], to drive efficiency outcomes for its customers.

According to GE CIO Jaimie Miller, the company intends to reduce its data centers by 90% over the next five years, a very aggressive target. The project is like “cleaning out an old closet,” says Miller, because you never know what you may find and you need to make hard decisions on what to refactor, what to rebuild, what to decommission. GE is also adopting off-the shelf cloud-based products from Salesforce and Box. To maintain security across these initiatives, GE “follows clear decision trees” on what to move to the cloud and when.“ 

Industrial Internet of Things and Big Data Analytics

GE’s Jamie Miller (left) outlines GE strategies for Cloud and OT / IT alignment to drive Industrial Internet customer success with GigaOm's Katie Fehrenbacher.

The impact of the Industrial Internet is evident across all industries. Take Florida Crystals, a leading sugar producer and distributor. The company now has sensors in its sugar fields and on its agricultural equipment. Big data and advanced analytics turn information from these sensors into actionable insights for the business.

According to GE’s Miller, the Industrial Internet requires a mash-up of IT (information technology) and OT (operational technology), two disciplines that used to exist in silos. GE’s brilliant factories model is an example of this trend: GE creates a “digital thread within the factory” -- from embedded sensors to MES to ERP and Sales -- providing visibility into how equipment is performing and how to optimize production. Miller characterizes the approach as “Lean on steroids with lots of data.”

Another interesting conference finding in this area: big data analytics is branching out from traditional domains, according to Amazon’s Vogels, with video and image analytics now growing rapidly.

New Cloud Architectures

Facebook’s Jay Parikh (left), commenting to Derrik Harris of GigaOm on Open Compute Project: “The network is the lynchpin of the data center.”
Facebook's Jay Parikh (left) and GigaOm's Derrik Harris

Facebook used Structure 2014 to announce the Wedge, its first modular network switch. The Wedge has been developed by Facebook as part of the Open Compute Project, alongside vendors such as Intel. Jay Parikh, VP infrastructure engineering, Facebook, said, “The network is the lynchpin of the data center, so we decided to build our own” to support project such as messaging, deep learning, and others. The classical separation of hardware and software is disappearing, according to Parikh: “We’re optimizing the full stack.” Still, he believes it is critical to keep things modular, to break infrastructure pieces into compose-able units (something he calls “disaggregation”), and to allow teams to change and adapt quickly.

This perspective was shared by VC investor Vinod Khosla of Khosla Ventures. “Something fundamental has changed, ” according to Khosla. Data centers used to be dominated by single vendors. However, traditional server architectures had been stuck for a long time. Vendors had no innovation for 30 years; in fact innovation had stopped, by and large, in the computer industry. That is, until Amazon, Google, Facebook, and other new players changed the game.

Khosla sees innovation in software-defined networks (SDN), software-defined data centers (SDDC), software-defined machines (SDM). “SDx is so powerfull,” he stated, “because the abstraction layers let you build applications on top, without the need to reengineer the underlying stuff.” The rate of innovation is increasing with the rate of abstraction.

Other speakers echoed this point of view: On-premise players will need to adjust both their architectures and licensing models as customers expect new consumptions models.


Finally: “40% - 60% of all IT costs are tied up in development and test,” according to Amazon’s Vogels, who makes the case that continuous integration, continuous delivery, and DevOps are the next forefront for Cloud deployments.