Today enterprises are allocating more and more resources in edge computing to accelerate the deployments of mobile applications and services, while boosting performance in user experience. As the era of IoT and IIoT is around the corner, edge computing can be observed in various environments, such as retail, factory floor, campus, public services, telecom stations, utility plants and transportations.
From Cloud to Edge
Most of the large-scale enterprises and organizations in the world have an intelligent cloud to deliver application and services to their customers and users. Some of them have even integrated A.I engines to improve their serviceability, for instance, global-scale financial institutes.
However, the widespread use of mobile devices instead of stationary systems has generated unprecedented volume of data that the cloud is encountered with bottlenecks to maintain the same level of serviceability, even as the cloud is constantly empowered to be more “intelligent”. Thus, it is time to make the edge “intelligent” to balance the load for the cloud.
At the dawn of 5G, artificial intelligence, machine learning, IoT, and VR/AR will be an ubiquitous phenomenon and in order to break through the bottleneck of the current Internet, it is essential to implement the cloud-like intelligence and analytics in edge computing to reduce the data routing time to the cloud. In short, the edge will have the functions of an intelligent cloud, like database, high-compute and analytics.
Concepts Behind Intelligent Edge
Intelligent Edge can be considered as an enhancement of the current orchestration of edge computing to address high-volume workloads generated by the devices at the edge, for instance, VR/AR goggles in factory floors, business applications at retail branches, and connectivity services for public services in multiple districts. Such workloads would cause latency if uploaded to the cloud. Thus, in order to boost real-time performance, Intelligent Edge is expected to process over 50% of total enterprise workloads once IoT is realized.
The idea of Intelligent Edge is implementing artificial intelligence, machine learning and/or even neural networks in edge devices, consisting of multiple white-box hardware, in order to run applications and services at the edge without compatibility or interoperability issue. In other words, some sophisticated applications that typically send workloads to the cloud will instead be processed at the edge. For instance, AR/VR applications at field service could generate tremendous workloads to the cloud and cause considerable latency. However, if the edge is empowered by A.I and machine learning like the cloud, such applications can be handled at the edge and users would receive the real-time results to improve performance and efficiency at the field.
Summarized Features of Intelligent Edge
- Address IoT workloads
- Balance workloads at service layer
- Real-time results in the field
- Connectivity at minimized latency
- Drive application and analytics performance
Recommended Solution
With all that said, the realization of IoT lies in the deployment of Intelligent Edge, instead of the cloud, because the edge device can address user data more efficiently than cloud server due to proximity to minimize the latency. However, in this increasingly “mobile” and “connected world”, the edge has to be “smarter” to process today’s sophisticated user/device data. In other words, edge device can no longer be an ordinary “edge device”, with characteristics of low power consumption, entry-level compute, storage and bandwidth. To run A.I and machine learning at edge, the white-box edge device has to be upgraded.
For example, LEC-2290 from Lanner is in a Box PC form factor, which is an ideal fit for Intelligent Edge environment considering the restricted space to fit in a rackmount server. LEC-2290 packs the needed compute and graphical performance to boost Intelligent Edge performance powered by Intel® Core™ i7-8700T/i7-8700, 2x DDR4 2133/2400 memory sockets, and the CPU’s built-in GPUs and accelerators. In fact, LEC-2290 has been showcased to demonstrate video analytics in an edge setting.
To further address capabilities in A.I and machine learning, LEC-2290 comes with support of Intel® Distribution of OpenVINO™ Toolkit and Movidius inferencing engine to realize high-performance deep learning applications.
Other critical features of LEC-2290 include 2 x RJ-45 LAN ports, 4 x PoE ports, 6 x COM ports, and 8 x DI + 8 x DO (DIO). For storage, LEC-2290 provides two removable HDD/SSD drive bays with RAID, 1 x mSATA socket and 1 x PCI Express x 16 socket. Regarding wireless connectivity, LEC-2290 is built with 1 x mini-PCIe (PCIe + USB signals) with one nano-SIM slot, and 1 x B Key M.2 socket (PCIe + USB 3.0 signal) with one nano-SIM slot as well.