Is connectivity the new bottleneck for edge data centers?

March 29, 2022 | Yesterday, I attended the first day of Data Center World in Austin TX. I have to say, I love having conferences in my home town - no flights, no hotel, no Uber. Just sleep in my own bed, have breakfast and head downtown :)

iGR looks at the data center market from the communications perspective. That is, we look at data centers, and especially edge data centers, as a point of connection for mobile networks and fiber. We are less concerned with the operational aspect of data centers - cooling, power, staffing, security, etc. So attending the conference was interesting for me to see the data center opportunity from their perspective, rather from that as a communications analyst.

One issue that came up quickly in the keynotes and workshops was ‘connectivity'. Data centers are obviously connected to each other and to major public cloud providers; these networks and solutions are well-established and this appears to be a thriving market. We will have more on this in the coming weeks.

But the problems arise when it comes to edge data centers and connectivity. Bill Kleyman of Switch presented the opening keynote and presented some interesting data from a recent AFCOM survey: a majority of enterprises said they were repatriating workloads from the cloud to on-premise data centers and that many companies planned to significantly increase the number of data centers they have.

Put this together and it appears that investments in edge data centers are increasing. But one of the very first slides in the next workshop (‘Connectivity for the 4th Industrial Revolution’) highlighted a major area of concern for ‘distributed data movement’ in the enterprise – connectivity. Connectivity was highlighted as a ‘major constriction point’.

Conversations with presenters and attendees during the workshop highlighted the frustrations the data center world has with the telco world – getting connectivity where they need it and when they need it. This applies especially to edge data centers and even those that are located at tower sites. Interconnect between large data centers can be provisioned very quickly (less than 15 minutes in some cases) but getting a connection to an edge data center can take months.

We have said many times before that the success of 5G depends on edge compute; we simply need to distribute the processing to lower latency times. And the success of edge compute obviously relies on being able to deploy that data center capacity in the correct location, securely. But it also means being able to connect to that edge data center with sufficient fiber to handle the workload. From what I heard yesterday, there is some way to go before this can be accomplished reliably to scale the edge data center opportunity.

---