Get started Bring yourself up to speed with our introductory content.

Flash makes IoT analytics descriptive, predictive, prescriptive

The Internet of Things has brought promises of industry transformation through the collection and analysis of vast quantities of data to improve the interrelated computing devices, machines, objects, animals and people. To undertake this grand connectivity scheme, cloud-based services are starting to mature and reduce the skepticism that IoT can flourish in the cloud.

Cloud must be the backbone for IoT. Cloud service providers (CSPs) are positioned to provide the operationalizing of IoT to help foster innovative, ISV-based solutions, and the storage, analysis and security of IoT data.

Today’s cloud architects face new challenges to design robust infrastructures that are built to withstand the persistent processing, bandwidth and storage demands of IoT in action. This requires end-to-end solutions that deliver efficient and reliable results, at scale.

The need for flash storage for IoT

This is where the promise of flash enters the equation. Flash memory can be found at many of the critical touch points of IoT. Without flash on the edge, mobile devices and other IoT end points wouldn’t be able to achieve nearly the same computing performance nor store the volumes of data essential to the analytics process. But the impact of flash doesn’t stop there. Flash inserts itself again at edge data processing locations where analog-to-digital conversion is performed. Flash is also a critical fixture in the centralized datacenters that support cloud infrastructures.

In a recently released report focused on how flash memory is driving new cloud-based applications, IDC in conjunction with SanDisk, discussed the complexities of the IoT ecosystem and in particular how these compute and data processing intensive systems must support storing and analyzing large volumes of data. The report calls out the importance of flash memory in IoT endpoints for “longer life cycles, temperature variations and the support field programmability.” To provide maximum value, “flash memory must be industrial grade, reliable and secure, and provide as much endurance as possible” and “afford immediate insight and control of IoT devices and processes.”

IoT data analytics require centralized datacenter infrastructures, i.e., NoSQL databases for persistent data, IoT archival data within data lakes and event stream processing. These infrastructures will be configured in distributed and cluster-based application architectures where the scaling of compute and storage resources may be required to scale independently; a cloud service provider stands to gain CAPEX and OPEX efficiency by taking a disaggregated rack approach for IoT services.

Flash has been identified as best at the performance tier of these datacenters where it can help deliver analytics on the volumes of IoT-generated data. Flash storage is necessary when accounting for the massive concurrent small file reads on a NoSQL database persistent store, along with the archival storage of historical IoT data in a data lake. Flash storage helps enable cloud-based IoT services to perform at scale and still meet the performance requirements of NoSQL read and write I/O along with timely Hadoop batch job processing.

What’s next?

Undoubtedly, the IoT space will become even noisier in the years ahead. In its 2015 third platform predictions, IDC estimates that there will be 200,000 new IoT applications and services, and 30 billion devices by 2020. As more and more “things” are brought online, businesses will continue to find unique ways to capitalize on new service models and benefit from real-time customer insights.

Though storage and analytics will, in some cases, move more to the edge devices when necessary for data to be closer to the point of collection, the deeper analysis and computation involved with more complex IoT applications will continue to task CSPs in meaningful ways.

The game’s afoot for cloud architects. For those who want to stay competitive in this bourgeoning and combative market, they’ll need to provide customers with the ability to use cloud infrastructure for IoT descriptive, predictive and prescriptive analytics. Otherwise, they’re at risk of being sidelined.

This article expresses the views of the author and not necessarily that of his/her employer.

All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.