Two concepts I’d like to add to Tom’s excellent thoughts here:
First: the computer product business (whether that’s servers, storage, networking, etc) has been centered on technology since the dawn of the computer age roughly 50 years ago. That technology was then “thrown over the wall” to be used by application software and customers. It was the enterprise sales process which smoothed over the gaps between technology and use. What we have seen over the last decade in hyperscale is that a handful of applications are becoming dominant, and the products subordinate to those applications. (Yes, it’s possible to debate in depth how this is different at Google vs Facebook vs Baidu vs Tencent vs AWS… but the point is the balance of power has shifted and the enterprise sales process is disintermediated in the new model).
Second: “Mass customization” is an important concept to bring to this discussion. Part of what made Symmetrix so successful 20 years ago were the myriad port modes that allowed the Symmetrix to work seamlessly with a wide variety of server models, each of which had been designed assuming it was the center of the universe with complete ownership of all its storage and other peripherals. The cloud (or edge computing, or whatever it becomes) will need to mass customize facing IoT devices if we are to maintain a reliable and secure Internet. A fully functional Ethernet connection — or even a fully functional wireless connection — does not contain errors, much less malicious code. An application aware connection (crossing server, storage, network boundaries) could.