Hadoop offline computing meaning

Question

I have been reading into Hadoop and one thing I have read a few times is that Hadoop is good for offline computing. What does that mean "offline computing"?

Answer 1

In on-line computing, user will wait for response to his query and expects the response in real-time or near real-time.

In offline computing, user will trigger/schedule an operation at a specific time and does not expect the response in real-time. Once the operation completes, he will check the result after sometime.

offline-computing will be done on huge sets of data, which may take hours to complete job and hence user is not interested in real time response. He will fire & forget the computing operation. He will check the results at his own time.

Answer 2

The terms "online" and "offline" are sometimes used synonymously with the terms "real-time" and "batch" respectively. For example, "online machine learning" refers to learning that occurs continuously as new training data becomes available. By contrast, "offline machine learning" refers to learning from a batch of training data.

Hadoop is good at processing large amounts of data in batches (typically executed on a relatively infrequent basis, e.g. hourly), as opposed to processing data in real-time as it becomes available.

Answer 3

AFAIK, "offline computing" is exactly what it says. It's a system disconnected from the web (usually for security reasons). But it's also related to dedicated servers that can access local databases, so they don't really need to connect to the web and neither they want to waste any resource.

So I think the document you've read was meaning that Hadoop is great for dedicated servers that need to process long running dedicated calculations, where CPU is a must. In this case, "offline" is usually related to performance, more than security I guess.