Non-disruptive addition of additional nodes through the same management interface and the ability for the same data to grow without moving it to the new hardware
we used to scale-up (grow until you reached max allowed configuration), the challenge of our day is distributed architectures, which leverage scale-out deployments. smaller components that are highly scalable.
in some ways it's more easily defined to people in "negative terms". That is, are you used to a dual controller system? Are you used to having to purchase more CPU/RAM capacity in a storage array upfront so can add disk later?
scale out = anything that can incrementally grow and in turn shrink. Things like infrastructure, apps, and my waist line are all things that can scale out
The term applies in multiple areas - hadoop clusters, general storage applications, hyperconverged (Nutanix, etc.), even backup and recovery more recently.
Also applies to cloud resources in that most cloud providers have written code to make all their internal resources scale-out. Customers don't see that but assume it's the case.
when talking "cloud" and diving in to the elasticity and flexibility characteristics, often there's some scale-out technologies under the covers to enable that
Anymore it seems like if we are talking scale out as opposed to up due to a change in application demand. Micro-services do not require big fat servers anymore
@johna_white the "shrinking" part is very interesting - it's often assumed but not always the case. Sometimes it's technically possible but not sufficiently tested/QA-ed for real world use.