Post written by
Gina Longoria is a Moor Insights & Strategy senior analyst for servers
Everyone is talking about the Internet of Things (IoT) and developing a strategy to go after this market. There is a significant amount of focus by many companies to create more “things”—smart devices, wearables, and intelligent industrial endpoints. However, IoT also creates a large opportunity for datacenter infrastructure providers to help their customers move, store, process, and analyze the data generated by the “things” and turn that data into relevant business insights. Although data is being collected at exponentially-higher rates than ever before, most organizations do not yet have the appropriate skill sets or systems to take full advantage of the data collected. Companies like Cisco Systems Cisco Systems, Dell Dell, IBM IBM, Intel Intel, Hewlett-Packard Hewlett-Packard, and Huawei all see this as a big opportunity to help their customers navigate these unchartered waters and to provide the datacenter infrastructure required to help transform massive amounts of data into actionable outcomes.
Moving Beyond the Cloud
Big Data analytics has become prevalent with many enterprise customers. The typical model involves gathering data from various sources, ingesting all of that data into the cloud datacenter, and using analytics engines within the datacenter to turn that data into insights.
This model breaks in an IoT world. Some estimates say in 2020 there will be over 200 billion connected devices with the amount of data being collected doubling every two years. With this huge number of devices and amount of data, it is not realistic or cost-effective to rely on data analytics strategies that are solely dependent on the datacenter. This creates the need for smarter and more robust compute capability at the edge of the network to allow for analytics closer to the endpoints.
Edge Computing pushes applications, data, and computing power away from the cloud to the logical edge of a network. Having compute capability closer to the endpoints significantly decreases the data volume that must be moved and the distance the data must go. Compute capability closer to the endpoints reduces transmission costs, shrinks latency, and provides real-time analytics capability to drive timely decisions. Distributed Edge Computing nodes must be managed and secured as an extension of the cloud-based infrastructure, and will potentially communicate with one or more cloud datacenters. Cisco Systems estimates that up to 40% of IoT data will be processed in the Fog (Cisco’s term for distributed computing infrastructure for IoT) by 2018.
There is Always a Tradeoff
The most helpful way to look at the Edge Computing concept is to examine specific use cases. In certain cases, it may be best to move towards a localized or partially-localized resource; in other circumstances, it might be best for all analytics to remain in the cloud datacenter. This all depends on the types of data being analyzed, the amount of data that needs to be managed, the associated connectivity requirements, cost considerations, and what specifically you are trying to do with the data.
Edge Computing is generally useful when you want to get information back as quickly as possible and where local processing power is needed to parse through and analyze data. Industrial automation, transportation, and any sector where webs of sensors or actuators are required may be a good opportunity. Use cases that are time-sensitive (like mobile healthcare applications) and sensor-rich environments where small time lapses can be very costly (such as the oil and gas industry) may also be a good fit.
In many cases, Edge Computing is not only used for real-time analysis but also used to parse out data to send to the cloud. This data is then married with data from other sources for further processing and Big Data analytics. For example, in a retail environment, you might want to use Edge Computing at each retail store to manage inventory, push in-store promotions, or deploy smart-shopping applications. In addition, for supply chain management and customer behavior analysis, you may want to gather data from multiple stores in various locations and send that data to the cloud to analyze trends that will help inform business decisions.
Edge Computing is not likely a good fit for personal wellness devices, home automation, and remote health monitoring due to the cost involved of providing intelligent nodes in home environments. In these cases, cloud-based applications will continue to be a perfectly acceptable approach for analyzing and compiling data.
It Is Still Anybody’s Game
We are just scratching the surface of the IoT opportunity, as the vast majority of industrial and consumer endpoints are not yet connected. Enterprise customers are just beginning to look at ways to enhance their analytics infrastructure to fit with these new demands of IoT data.
Each datacenter infrastructure vendor is developing IoT strategies and products with their own perspectives on the market based on where their core product strengths lie today…whether from a sensor / endpoint point of view, a network point of view, an analytics software point of view, or from a cloud datacenter point of view. And as this market matures, I don’t believe there will be one dominant vendor with the winning solution to meet the wide range of needs for IoT analytics that span from sensor to datacenter. Vendors that will remain in the fight will be those able to demonstrate they understand the needs for specific vertical markets and can deliver solutions that help enterprise customers save money and improve their ability to turn IoT data into insights.