DataCentreNews UK - Specialist news for cloud & data centre decision-makers
Story image
Panellist insights on edge and hyperconverged architectures
Fri, 8th Mar 2019
FYI, this story is more than a year old

I have a confession to make.

I have come to realise that I am size-ist as it relates to data centers.

My history in the cloud - service provider space began on the power side of the world. To me, hyperscale has always been the data center market to me. The edge opportunity, however, is opening my eyes to my own prejudice, and the panel that I had the chance to see at Edge Congress only enhanced my excitement.

Panellists talked about the dramatic spike in data that will need to be handled at the edge, vertical-specific edge requirements, and major implications for telco edge.

Lastly, they covered the critical success factors in deploying an edge architecture which included a nod to a hyperconverged solution.

Edge Congress Panel – A discussion on opportunities and challenges at the edge

The panel featured four experts who are on the front line helping customers implement edge computing solutions every day.

They included Paul Morgan, Global Sales for Manufacturing, Automotive - IoT, at HP Enterprise (HPE) and Aad Dekkers VP from Scale Computing, which makes infrastructure specifically for micro and nano-data centers.

Also, the panel included two of my colleagues from Schneider Electric, Cyril Perducat, EVP IoT - Digital Offers and Joris Verdickt, Segment VP, Enterprise End Users in the Secure Power Division.

Of the key highlights from the panel, the following two quotes from Paul Morgan really capture the opportunity:

“While today only 10% of all data is handled at the edge, analysts expect in 3 years between 50% and 75% of all data to be produced and processed at the edge,” said Morgan.

“Gartner puts the figure at 75%,” he said, noting it will come from applications such as video solutions producing petabytes of data per day while others pointed to autonomous cars.

“Edge and edge computing are going to be key moving forward, because of the huge amount of data being produced,” Morgan said. “Bandwidth is going to be absolutely key.

With that as a backdrop, the following are my takeaways from the panel discussion around the challenges and opportunities at the edge.

Edge computing requirements vary by vertical

Industrial companies such as manufacturers typically have more complex edge requirements, because they must fit in with lots of legacy infrastructure.

The retail sector is typically less complex but puts a premium on availability and driving cutting-edge technology to brick and mortar retailers to combat their online competitors.

These kinds of demands are all impacting retail IT and the edge environment. While 27% of IT managers in retail say they're prepared for what's coming, 20% are holding off on new applications because the IT infrastructure isn't yet where it needs to be, Verdickt said.

Available infrastructure for telco edge is promising

Telecom companies have some of the most demanding mobile edge computing requirements, and also some of the most mature edge infrastructure.

Having already spent billions on infrastructure, both in full-blown data centers and smaller points of presence located practically everywhere, telcos are already handling the loads that bandwidth-intensive applications such as video presently. But the job is about to get more challenging for telco edge. Consider connected and autonomous cars, which Morgan said produce up to 40 petabytes of data per day.

“That will all need to be offloaded and telcos are positioned for that,” he said.

What's more, it all has to happen in real time, noted the panel moderator, Perducat. This is no small task, but telcos' delivery of reliable solutions and connectivity is a marker that they can succeed in powering critical, time-sensitive applications.

5 pillars for success at the edge

No matter the sector, Verdickt said he sees five pillars for success at the edge:

  • Availability of the network
  • A clear strategy for capturing data, such as use of Bluetooth in a retail environment
  • A centralised model for IT staff with a broad-based management platform for controlling edge environments
  • Attention to security, both physical and cyber
  • Use of converged infrastructure

Use of hyperconverged architectures at the edge

Panellists generally agreed that it was important to keep edge architectures simple, given there's typically little to no IT staff on site at edge locations.

One way to do that is to use hyperconverged infrastructure, where all compute, storage, and network components are delivered in a single enclosure, complete with software to operate and manage it all remotely.

Such a strategy can also help keep costs down, said Dekkers. And it's important that the infrastructure be able to not only scale up, but down, he noted.

“There are certain parts of the edge that require less in terms of number of virtual machines or number of applications. But still they need to be up,” he said.

The idea is to deliver the same data center-type functionality but in a much smaller footprint.

Planning for the (edge) future

With the anticipated, dramatic growth for edge computing, panellists addressed how to plan for it.

“We will need different ways to compute and process [data] moving forward,” Morgan said, which will require continued innovation to develop new compute platforms, and in a way that takes the burden off IT in managing the environment.

He envisioned a service model developing around edge data centers – “edge as a service,” as he put it.

Perducat agreed, noting the market needs a combination of the right hardware, software, and services at the edge to not only create new experiences for customers but to do so in a simple, efficient way.

He suggested we should look at the edge in a new context and create different innovative use cases that use the cloud as a complementary resource.

Check Out the Edge Congress Panel Discussion  

That is my takeaway for the edge panel discussion, but view it yourself for additional insights.