Genius Edge – Insight to Action at the Edge

By Pradeep Arora

To be successful an enterprise needs to outperform its competitors and digital transformation (DX) of business processes is the way to go about it. The edge is the place where data is generated by the “things” and consequently a lot of “internet of things” innovation is taking place. Now every vendor out there is offering to help you build an “intelligent edge”. We at HPE advise you to build a “genius edge”, not just a simple intelligent one. A genius must do things better than others and it is best to leverage existing technologies and processes that run your business, adapting them as needed. Corporate IT has long standardized on compute, storage and networking and the cost plus efficiencies have kept improving drastically. The DX is leading to OT-IT convergence of both human and capital resource. The edge needs to be enabled using standard data center components coupled with methodologies of security and remote management. 

That is the beginning of a genius edge, a real genius will soon realize that why not put software I want to run at edge on the same pedestal. My IT software should run as-is on the edge for OT – why do I need 2 different versions each with its own lifecycle challenges. Now I can build my AI models in standards base data center or in the cloud and then these models right on the edge. Wow! Before we get too far please realize that OT environments tend be harsher than data center and then there is need for an optimal way of running software – “Balanced workloads across the paradigms of edge, core and cloud”. With this mind HPE has created a brand new category of “Converged Edge Systems” that coupled with our entire portfolio plus our partner ecosystem will make your edge a “real genius”. 

Please come and see our Dr. Tom lead a panel discussion at HPE Discover with our thought-leading customers Tesla, Comcast, & Alfa Romeo Sauber F1. PNL4857, Tuesday, June 19th, 12:00-1:00 pm 

Intelligent ERP – Core to Edge
There are two sources of data in an enterprise, the traditional ERP systems and the source of a huge amount data at the edge. Greater business value can be harnessed by taking edge data to the core / clod and by taking ERP data to the edge. This is done at or near the edge by running analytics on “data in motion” in an ASAP manner because most of the data has lower value over time. Please see picture below for 7 reasons one must compute at the edge and associated 7 pitfalls one must watch for. 

What a lot of my customers are not aware of is the concept “Big Analog Data” illustrating the fact that most edge data is in analog form.  It is not on the wire and mostly unharnessed for its potential benefits. If a model you plan to use runs only in the cloud based offering there may not be enough bandwidth, time window for processing or even more so, the funding to do this. 

On the other side of the ERP data, extensive models can be built in the data center and then run at / near the edge.  Now we have opened the door for four fundamental efficiency mechanism within the organization namely, inter unit, intra units within the region, Intra Company across my business units and finally as competing with everyone in world. The ERP models look at larger geography and longer time period data and collapse them into a business value creating runtime for the edge.  We take this to the edge and get benefits from the edge data ASAP. Discover blog graphic.png

The edge processing deals with incoming data and we can categorize that into “good”, “bad” and “ugly”. Good means all is well and bad in when models detect an anomaly. Ugly is that our models do not know what to do and we need different parameters for the model or possibly a different model altogether. Thus the need for a CI CD deployment pipeline being implemented to allow for quick changes to be created and then tested in a limited number of sites. Final approval deploys them widely using automation. The ugly are now good or bad again. Without this cycle of improvements we would suffer from model drift and non-alignment with data shifts that will happen over time. 

At one of our customers, SAP applications are generally written using Java with 32GB being maximum memory available. They are working on taking advantage of ERP data by creating new insights for the edge via models that are using neural networks being created on data center equipment. The runtime will happen on HPE EdgeLine Converged systems. The Cloud Foundry will manage the EL4000 nodes on the edge via Kubernetes allowing for CI/CD agile development. A model once proven at one site will be deployed to 3 others for blue green system of limited testing. The final model will be deployed at every site. The same system is also being designed to allow a different data pipeline for taking repair manuals to the edge site. Final plans are for automated repair procedure using vision goggles. 

Elastic & Agile -- Cloud Agility with On Prem Control

Becoming a digitally transformed organization also asks for an agile and elastic organization. Efficiencies obtained by an AI model at one of the edge sites needs to be replicated ASAP to all other edge sites. When you standardized on IT procedures you also chose their purchasing policies which can be rather long. Our models may need to run on base operating systems, in a virtualized environment or within containers. Our choices must enable this fundamentally. Our answer is simply stated as “Under what kind of consumption model would you like hardware, operating systems, applications along with the control plane? “, worldwide from a single vendor – of course.


0 Kudos
About the Author


HPE Alliance Partners