Cloud computing is a long way from being fully mature, but its obsolescence may already be upon us. Is the cloud’s future really up in the air?
As Peter Levine, a partner at venture capital firm Andreessen Horowitz, puts it, “Everything that’s popular in technology always gets replaced by something else,” be it Microsoft Windows, minicomputers exemplified by Digital Equipment Corp., specialized workstations typified by Sun Microsystems, or, yes, even cloud computing.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
As Levine explains it, cloud computing, which he views as the centralization of IT workloads into a small number of super-mega-huge datacenters, is an unsustainable, unworkable, slow-to-respond method. The need for instantaneous information makes the network latency associated with a device-to-datacenter model and the corresponding datacenter-to-device return trip simply too long and therefore unacceptable.
Computing, Levine suggests, will move to a peer mesh of edge devices, migrating away from the centralized cloud model. Consider smart cars. They need to continually exchange information with each other about immediate, hyperlocal traffic conditions. Smart cars need to know that an accident occurred 10 seconds ago a half-mile up the road, that a pedestrian is entering a crosswalk, or that a traffic light is about to turn red. For this to work requires realtime data collection, processing, and sharing with other vehicles in the immediate area. The round-trip processing in the cloud model isn’t even remotely (pun intended) fast enough.
Text messaging is similar in that messages exchanged between people sitting just feet apart are still routed through a distant datacenter. It’s inefficient, slow (in compute terms), and unsustainable. The centralization is needed only for logging and journaling.
The answer, Levine postulates, is pushing processing and intelligence out to the edge, using many-to-many relationships among vehicles for information exchange, along with edge-based processing based on super-powerful machine-learning algorithms. No wonder he describes the self-driving car as “a datacenter on wheels.” Similarly, a drone is a datacenter with wings and a robot is a datacenter with arms and legs. They all need to process data in real time. The latency of the network plus the amount of information needing to travel renders the round-trip on the cloud unsuitable, though that’s still plenty fast enough for a Google search, he says.
The cloud still plays a role; data eventually needs to be stored, after all. That makes this model not fully edge and not fully cloud. It’s perhaps closer to what Cisco dubs “fog computing.” It also speaks to the inevitability of how IoT-driven smart cities must operate, a concept explained to me by Esmeralda Swartz, vice president of strategy and marketing at Ericsson.
There’s a profound irony to this. We started the age of IT (MIS as it was then known) with the IBM mainframe as the centralized place where all programs ran, all processing was done, and all data was stored. That was blown apart by decentralization, driven by the client/server model, Ethernet (or Token Ring or ARCnet), network operating systems (NetWare, VINES, LAN Manager, 3+ Open, Windows for Workgroups, Windows NT, OS/2 Warp, etc.) and early network-aware databases, such as Btrieve. Cloud computing swings the pendulum back to the centralized data model of the past, albeit with a dose of edge processing.
It’s throwing out everything you know and seeing from a paradoxically different perspective — just like the young girl presented on Christmas morning with her great-grandmother’s heirloom wristwatch, only to declare, “A watch that doesn’t need batteries? Gee, what will they think of next!”
You can watch Levine’s presentation “Return to the Edge and the End of Cloud Computing” on YouTube.
No doubt you’ve already thought about this. Where do you think cloud computing is headed? Is this a technology that is ultimately doomed to be superseded by something different, better, faster, and cheaper? What does this mean for you as an application developer? Share your thoughts and fears; we’d like to hear from you.