Data centers, by their very nature, have been centralized affairs. Instead of having a server in every office, or a small server room on every floor, it is far more efficient to have one data center to serve the entire building or campus.
Taking that a stage further, it is far more efficient to have one data center serving regional, national, or even global operations. Of course, the needs of disaster recovery, resilience, security, and latency often dictate the need for several large data centers — but they still represent a centralization of IT infrastructure.
Those days may be over. Edge data centers are all about distributed computing, taking the compute power or the storage or networking to the point where it is most needed. Self-driving cars need edge data centers all along the highway and at intersections if they are even going to be a success. New telecom services such as 5G demand so much power that they can’t work well when everything is fed to a central data center.
Here are some of the top trends in the edge data center market.
Traditionally, an edge data center was simply a slimmed down replica of the full data center – with the attendant physical security, power, and cooling challenges.
Not anymore. Satish Ramakrishnan, a senior executive at MinIO, thinks the edge data center is going through a miniaturization process, one that will continue for the foreseeable future.
“The new edge data center is software-defined, runs on commodity hardware, can scale to double digit petabytes, and treats bandwidth as the highest cost,” said Ramakrishnan.
“Given the density of modern NVMe drives, the footprint can be the size of a utility box with limited cooling and power requirements. It can be mobile and can support multiple backhauls options.”
Ramakrishnan with MinIO mentioned the software-defined aspect of this. Software-defined technologies are playing a critical role in the attainment of edge data center objectives.
Although edge data centers are shrinking, their functionality must match that of a far larger data center.
“It places a greater emphasis on the efficiency of software-defined infrastructure, requiring portable containerized infrastructure to support cloud-native applications,” said Ramakrishnan.
For example, MinIO’s binary is less than 100 MB, of which half is the graphical console, and still provides fast object storage at the edge. This means containers can be packed densely and orchestrated more easily. This is the same code that runs in the data center – ensuring performance consistency from edge to core.
3. Data center anywhere
Where is all this heading? Instead of the workloads and data being sent to a central data center, the opposite is becoming the case.
“The net effect is an architecture that supports a data center anywhere mentality through software-defined infrastructure,” said Ramakrishnan with MinIO.
“Expect this to push the traditional edge data center closer to the actual application workloads – from autonomous vehicles, industrial internet of things (IIoT), or even gaming applications.”
4. 5G Necessity
Edge today means many different things to different people. But it is gradually becoming more focused. Why? It is largely being defined by the needs of 5G use cases.
“The needs of 5G can be summed up as density, sustainability, low maintenance, and perhaps the most important one – low complexity,” said Hao Zhong, co-founder and CEO of ScaleFlux.
“These are the aspects by which edge storage products will be compared to each other.”
5. More density, less power
Edge data centers are getting smaller while being able to pack more punch per square meter.
But the demands of 5G and other edge applications require vast changes to IT infrastructure and IT thinking in general.
“We all know that to prepare, we will have to learn how to sort, store, and process more data closer to the edge, but we haven’t thought through what technology we will have to deploy to accomplish that,” said Zhong with ScaleFlux.
“The inconvenient truth is that we will have to do a lot more with a lot less; much higher storage density; working in tighter operating envelopes; reduced power budgets and cooling; running for longer life cycles with less maintenance; and analyzing more data with more compute stuffed into fewer nodes.”
The power requirements are particularly striking. 5G and these other edge services require wonderful performance. Users expect edge-based cloud services to deliver sophisticated and customized and localized services with the performance you would typically expect from a colocation provider. Due to grid constraints and ongoing environmental pressure, this must be accomplished without too much power drain. Take the case of storage infrastructure.
“Ordinary storage technology will not succeed in these conditions; edge storage solutions must evolve to address some of the challenges of traditional server architectures that we’ve lived with for too long, like performing all storage processing in the CPU,” said Zhong.