The cost of local versus remote computing often varies volume. At low volumes, remote computing often is more affordable in terms of total cost of ownership. At high volumes, local computing often makes more sense.
This example of the electrical energy requirements for computing costs--done locally versus remotely--illustrates the principle.
The same sort of trade off often exists for latency-dependent apps, where the time to analyze data and then cause a system response is stringent. In such cases the trade off is not financial but performance related: distant computing might simply be inadequate because of total network transmission delay between a local controller and a distant computing resource.
The point is that, to a large extent, communications and local computing are essentially substitutes for each other for a computing solution. The communications enables remote computing (hyperscale cloud computing, for example) and now also some forms of edge computing.
You can see that in this chart from IBM of various computing functions that are virtualized by the use of remote functions.
In principle, such trade offs are embedded in choices made between network slicing and private networks as well; or wide area private networks and local computing; or edge computing versus remote hyperscale facilities use.
The point is that the physical siting of computing resources allows use of communications networks to support remote computing as a driver of computational cost, when app performance requirements do not dictate the choices.
That contributes to the range of drivers transforming traditional “telco” networks into “data” networks” and making communications a part of the computing value chain, just as communications networks now are part of the internet ecosystem.
No comments:
Post a Comment