The obvious positioning for any firm involved in cloud computing (using hyperscale or other remote facilities) is that edge computing works with cloud computing. That’s true.
“Edge computing doesn’t compete with cloud computing, per se, but it will complement and complete it,” say analysts at Gartner. Edge computing addresses the limitations of centralized computing--such as latency--by moving processing closer to the source of data generation, “things” and users.
Of course, there are many “edges.” And neither deployment at the edge, nor IoT always and necessarily require edge computing. Some edge computing tasks can be completed right on a device. PCs, controllers or perhaps even autonomous vehicles,
Some operations only require access to a remote data center, such as when smartphones consume content stored at hyperscale data centers. That, again, is an example of “edge” devices not requiring edge computing.
The unique driver of edge computing is latency performance: when neither the local device itself or a remote data center can supply computing with required response times. But other use cases exist.
When huge amounts of raw data are produced, it might be quite useful to process locally to avoid WAN bandwidth investments. The trade-off there is the cost of local processing compared to the cost of WAN bandwidth.
In other cases, when network connectivity might not be assured, it might be useful to allow uninterrupted operations at a local level even when WAN connections are periodically disrupted.
In other cases there could be privacy or security concerns that create incentives not to send data across WANs. In yet other cases legacy computing resources must be integrated.
The point is that the need for edge computing is different from device computing at the edge. And edge computing is different from infrastructure edge computing.
Sometimes enterprises will host their own edge computing on the premises. In other cases apps might require access to multi-tenant infrastructure edge computing resources. Such needs will develop over time as use cases emerge.
source: Gartner
No comments:
Post a Comment