Orchestrating a Symphony of Data
The Internet of Things (IoT) creates a sea of data across networks, protocols, and devices. The piecemeal capture, normalization, and analysis of data will over time be impossible at scale unless analytics move to the edge and will only be effective if the computational environment is intelligent, context aware, and completely agnostic towards the underlying connectivity technology. Smart solutions will require development, not just of good analysis and logic but also of ‘soft sensors’ (virtual sensors) that can live inside and dynamically assemble across a common, resilient, smart platform to effectively tackle the complexities of the IoT.
VizLore, which specializes in multi-device management solutions, combines a potent blend of technologies to develop an agile IoT platform that creates and manages soft sensors across networks and devices and uses them to perform data analytics in an adaptive and scalable fashion. Soft sensors are elementary data analytics modules that are dynamically assembled into a distributed IoT data analytics platform to make data actionable. Examples include optimization, controlled operational processes, and workflows in a predictive and preventative environment.
VizLore has created a platform with real business benefits that internalizes complexity to provide ease of implementation and future-proof extensibility to users.
Data in the Digital Economy
In 2013, it was reported that a full 90 percent of the world’s data had been generated within the previous two years. At the rate of 10 times more data every two years — a trend that has held true for the past 30 years — the same report today would state that 99.5 percent of all data has been created within the last five years.
This data comes from everywhere: social network posts, digital pictures and videos, banking and commercial transaction records, smartphone Apps, and sensors used to gather operational and contextual information. We are in the midst of a major technology revolution, and digitalization is now dominating every sector of the economy. This global trend accounted for 22 percent of the world’s economy in 2015 and is expected to reach 25 percent by 2020.
These volumes of data make up what has been termed ‘Big Data.’ Data analytics are a fast emerging technology that creates business value by combining and analyzing past data sets to predict problems and propose actions for their mitigation.
Two Sides of the Same Coin
There are two kinds of Big Data use cases: ‘Data-at-rest’ scenarios involve historical data safely stored and easily accessible, and ‘fast-data’ scenarios that involve tapping into the value of data in motion. The IoT is more about ‘data in motion’ in which the mission is to capture and extract contextual value in real (or near-real) time.
Fast data is front and center in the world of IoT connected devices. The IoT gathers unprecedented amounts of data from disparate sources and by correlating systems, data, and people — often powered by Artificial Intelligence (AI) applications — to create solutions that are capable of fundamentally changing the affected organizations.
IoT analytics can leverage into business benefits in three fundamental ways:
- Automates manual and error-prone processes, freeing up resources to focus on the most valuable parts of the operation
- Strengthens the relationship between the company and its customers, bringing a deeper, richer understanding of customers’ needs
- Upgrades the company’s conventional business model by creating new revenue streams and pricing strategies; upgrades individual, one-time product sales to connected services that generate recurring revenues
Two Computational Strata
The IoT is poised to change the way we currently live, work, and play, as we will soon be surrounded by all manner of connected devices, sensors, wearables, and appliances that will each come with their own set of functionalities. There is surely no guarantee that any two or more systems will know how to talk to each other or share data. To derive the expected long-term value from the IoT universe, VizLore is taking an integral approach to help make sense of it all. The way to do that is to implement a security-focused Software-Defined Networking (SDN) Platform-as-a-Service (PaaS) architecture that is designed to orchestrate the devices and flow of data across IoT networks.
The VizLore platform is physically divided into two computational strata: ‘Edge computing’ is applied and distributed across connected devices and networking elements to enable tactical processing of fast data and serve to improve service resilience; and ‘Cloud computing’ orchestrates distributed computing and service integration across the edge devices and performs strategic oversight and workflow control.
To improve the effectiveness of the IoT platform, VizLore uses an open, purpose-built SDN as underlying fabric on which the Edge platform rests. This SDN solution enables distributed data gathering and processing due to the following:
- Edge network devices, such as Wi-Fi, gateways, and sensors, require that SDN logic be closer to end users where response to specific events requires decisions faster than the cloud can process a response.
- The SDN control plane is used to dynamically assemble physical sensors and network performance data into soft sensor constellations.
- Scalable and versatile data analytics are achieved through juxtaposing soft sensors across devices, networks, and the cloud to provide actionable insights on system performance.
Keep IoT Data Flowing
When SDN is complemented with edge and fog devices, such as switches, firewalls, IP cameras, sensors, and Bluetooth beacons, it opens up a whole new world of possibilities for the creation of added-value IoT services. Since the software part of the network sits in the cloud, it can be easily updated and redefined to suit any future need and maintain security without the need of an IT person on premises. The fact that VizLore’s SDN picks up sensory data across the connected devices means that it can compile information from different sources into soft sensors that can then be further combined with other soft sensors to provide actionable recommendations.
SDN fabrics that are deployed edge devices and coordinated from the cloud are able to conceal the complexities of IoT network management so that users are able to focus on optimizing their critical business processes. SDN fabrics enable soft sensors to transparently blend many different protocols and data formats, including Wi-Fi, ZigBee, AllJoyn, and Project Haystack, which can be found in complex use cases. SDN fabrics are able assemble these disparate sources into soft sensor networks managed through the API abstraction layer that is exposed to our cloud-based platform.
VizLore’s IoT platform architecture is flexible and open, and allows various types of devices, sensors, appliances, and wearables to be inter-connected. Data is extracted from the VizLore system in ways that are ultimately useful to the end user.
Architecture of VizLore’s IoT Analytics platform
Soft Sensor Stack
There are two categories of soft sensors: 1) those optimized for processing data in motion, called DiM soft sensors; and 2) those specialized in processing data at rest, called DaR soft sensors.
DiM soft sensors can be deployed on network and edge devices to enable collection, formatting, and analysis of all available contextual data. Networking elements include sensor primitives that provide data such as interface status (queue levels, signal-to-noise ratio, Tx/Rx bit rates, and packet drops), system status (CPU/RAM load, local storage, routing tables, and power consumption), and environmental status (connected devices, active/inactive links, and surrounding Wi-Fi networks). By combining sensor primitives, soft sensors are equipped to present contextual results without needing to perform post-processing on dedicated servers or SDN controllers.
In general, this approach requires the deployment of a software agent on each network device (Wi-Fi access point, distribution switch, or network router). The software agents are built to bridge the sensor primitive handlers for the Operating System (OS) of each device, such as OpenWRT (Linux), with the data analysis model of the soft sensor. The agents provide an abstraction of OS-controlled resources that data analysis processes can utilize without further formatting or preparation. Agents speed up data analysis and enable different software sensors (different data analysis models) to be deployed on different networking equipment. The main components of a software agent are:
- Software hooks for integration to expose the status and measurements on networking interfaces and other networking node resources, such as CPU and storage
- Buffers for sensory and contextual data necessary for accurate formatting
- Data formatting procedures to expose sensory and contextual data collected from the underlying OS to data analytics procedures of DiM soft sensors
- Action handler modules of the software agents to transform results of prescriptive analytics into configuration commands for networking nodes
- Query handlers to provide API elements that enable soft sensors to obtain formatted sensory and contextual data from the software agent’s buffer
In SDN logic, a software agent is implemented either as a southbound API (if DiM soft sensors are implemented on the controller or networking device) or northbound API (if DiM and DaR soft sensors are deployed on the cloud or SDN controller with higher hierarchy).
There are different layers of soft sensors. DiM soft sensors are usually located on networking elements, and they are used for combining the analysis of measurements and inputs from sensor primitives, which is a function of the network element’s OS. DaR soft sensors are deployed on the virtual network controllers that are hosted on cloud platforms. These DaR soft sensors combine results provided by DiM sensors, sensor primitives, and other sources of contextual data, such as web portals and open databases. They provide the deeper analytical insight that is necessary for proactive resource and service provision planning.
The figure (opposite) shows how different sources of contextual data and sensory measurements can be combined to create a layered structure of DiM and DaR soft sensors. Each next layer provides deeper analytical insight into network status, trends, state transitions, and correlations between network events. Layered assembly of DiM and DaR soft sensors is enabled via a well-defined Representational State Transfer (REST) API (the API abstraction layer). This allows deployment of soft sensors and prescriptive and predictive IoT data analytics when and where they are needed (depending on the services and application on a particular communication network).
Many new modern services can be supplied through IoT analytics platforms, and the flexibility offered through a soft-sensor approach and two-strata architecture is designed to produce tangible business benefits. Adopters can:
- Expedite innovation, expand market reach, and improve customer experiences without necessarily increasing operational complexity.
- Continually monitor, analyze, and optimize development and manufacturing processes in support of optimized work flows and efficient operation.
- Manage a complex and ever-expanding portfolio of physical assets that are distributed across the globe and effectively supervise a growing ecosystem of supply-chain partners.
Sensors at Work
The challenge of detecting unauthorized wireless network routers has increased by orders of magnitude at a time when the verification of edge devices, sensors, and wearables is understood to be critical. In one ‘smart’ building deployment, VizLore has implemented ‘Rogue AP detection’ in the form of a composite soft sensor that detects unauthorized access points that might pose a security threat.
Users of IoT-enabled intelligent networks are limited only by their imaginations for building any new service or bringing old processes fully into the twenty-first century.
Stay Ahead of the Competition
Every forward-looking business has to develop new skills and organize its workflows around the IoT, Big Data, mobility, and cloud computing to preserve its competitive edge in this digital economy. To effectively exploit opportunities of the digital age, organizations must overcome the challenges of integrating data from multiple sources, automating the collection of data, and analyzing data to effectively identify actionable insights.
The VizLore soft sensor-based IoT analytics platform is well suited to address these challenges. We are prepared to equip both traditional and nontraditional players with the necessary means to address the disruptive opportunities that are forecast by these trends. It is important to remember that the IoT is about the data, not the devices. Data is the twenty-first century’s primary raw material — and as always, raw materials represent great potential value. Collect data responsibly, handle it efficiently, and learn to work with it on a broad scale. Do not over-plan: Face the future head-on and engage!