Adaptive, composable pools of compute – Gigaom Structure

Gigaom Structure

[My top-level notes from the”>Gigaom Structure conference…]

The big picture – affordable and easy
The Structure conference focused on the evolving territory of cloud infrastructure, highlighting some fundamental shifts in the industry. First, the enterprise has been challenged to overcome the cost, deployment, and management overhead of adoption. However, many emerging businesses are serving this need by making it easier to deploy and run these services. Now pretty much all enterprises understand the value of moving into either a private on-premise or public multi-tenant cloud (and there was much discussion about when to co-lo and when to go public). Adoption is further enabled by the price war between Amazon, Microsoft, and soon Google that has driven public cloud services to become more affordable.

… agile & elastic
The second big shift is in making networks more flexible, elastic, and agile. Services are now more easily deployed across abstraction layers like virtual machines, or modularized into containers. Both VMware and Docker had a strong presence at Structure and most talks had some refrain about the relative merits of one versus the other. Network hardware is softening or virtualizing altogether into SDN and NFV solutions. It’s much easier and cheaper to update software than it is to update hardware. In parallel, more machine intelligence is displacing both hardware and human IT resources, enabling efforts in self-optimizing networks (SON). All of this makes for networks that are sensing and responding to constantly changing conditions.

…composable pools of compute
Third, compute power has become a distributed commodity that is dis-aggregated, addressable, and composable from anywhere on the network. Hypervisors and containers become the means for addressing compute pools, with services stretched across these hardware-agnostic abstraction layers. Notably, there was much talk about how the Internet of Things will force a reconfiguration of networks as billions of devices come on line, some of which require very low latency for their control loops. Pushing compute out to the edges where it’s needed for industrial IoT will spare the core from being overburdened by compute requests.

The Big Picture is starting to show a world awash in pools of computation and heterogeneous networks that are becoming more intelligent and adaptive.

Comments are closed.