RECAP | ROUNDTABLE | How do you handle Edge Data Center Resiliency?

Edge data centers are some of the most widely mentioned topics today in the data center industry. In this roundtable, with the support of Tomas Rahkonen, Distributed Data Centers Research Director for Uptime Institute, attendees were provided the opportunity to engage on how they are addressing Edge Data Centers and Resiliency.

Roundtable attendees expressed an interest to discuss a number of edge data center topics:
• To learn more about edge data center resiliency
• Dealing with increasing workloads at the edge
• Sustainability and edge data centers
• How edge data centers are developing for different businesses
• Strategies on dealing with edge environments
• Ways to deploy and manage edge data centers

Tomas kicked off the discussion by saying that by 2025, 75% of data will be processed at the edge, projected by Gartner. Urban electrification makes power sources and supplies less secure. Weather emergencies and other potential outage causes are creating concerns over operating at the edge as well. Clouds have availability zones - does edge need the same? This raises the initial question how you define the edge.

How do you define an Edge Data Center in your environment?
The challenges at the edge relate to Networks, Power, Support and Application resiliency (legacy app’s not programmed for distributed deployment, having problems with security, application management, etc.). For a bank/financial company, Edge Data Centers are needed at locations of heavy trading. Supplementary forms of processing outside of the centralized sites where heavy transactions occur is what is defining the edge. Retail has many use cases for computing at the point-of-sale. Note that retail is likely one of the front-runners of distributed edge computing. An attendee from a manufacturing company mentioned “server rooms” used in factories as being the edge. One driver for Edge Data Centers is large data files related to automation and related data logging. Business analytics and related transfer of large files was also mentioned as a use case driving a need for Edge Data Centers.

In regard to retail, an attendee described the edge as being the storefront – the closest point to their customer. He mentioned how at the storefront, workloads have been extreme during the pandemic. They are looking to provide a great shopping experience where the customers are located. A lot of the growth is located at the population centers. Running applications centrally but providing a seamless checkout experience at the storefront edge has been a challenge. Bottlenecks have been power at the store level, and then network at the distribution center level. Power and network appear to be the two Achilles heals. Maintenance and repair are done through remote experts and outsourced technicians.

At the edge, do you handle resiliency differently than at your data centers?
Level of resiliency depends on enterprise risk appetite. It should be determined in close relation to the IT service level agreements (SLAs). An attendee mentioned they are focusing on how to use regionalized capacity at distribution centers to improve resiliency and performance instead of bringing everything back to a centralized location, and all the while retaining the ability to scale quickly while maintaining super high levels of traffic.

Do you deal with redundancy per site, geographically with multiple sites?
Edge resiliency surfaces as a critical point now with the weather-related emergencies including power outages. Many are in the beginning of their edge computing journey. Steps being taken include re-architecting of app’s, defining required levels of uptime and site resiliency, operation concept (lights out or degree of staffing). Legacy applications are not agile to move and in some cases cost-prohibitive to reprogram for a geographically distributed architecture. In regard to operating in a lights out environment, higher levels of local redundancy are being considered.

Are there new technologies that will be important for edge resiliency?
Regarding the use of LTE/5G to achieve a redundant network path to complement fiber, it was mentioned how in the recent extreme cold weather event LTE has shown reduced capacity. One needs to trace the networking upwards in the systems to see if LTE/5G gives a truly redundant network path.

How do you think edge data centers will be operated?
One attendee mentioned they are using remote experts today. Maintenance and repair done through remote experts and outsourced technicians for the most part. It was mentioned how the facilities team and IT team report to the same manager, which is important for coordination.

In summary, most appear to be at the beginning of their edge data center strategy journey, or as one attendee called it, the beginning of their workload placement strategy where workload placement drives the need. The implementation bottlenecks are around power, network, maintenance, and the architecture of applications utilizing the edge themselves. The appetite for risk is what’s important because it will define how you invest and deal with these bottlenecks. But overall, there still appear to be more questions than answers right now as companies seriously start to consider and deploy edge distributed infrastructure.