Heh, most systems could fit their entire working dataset in 1TB and never worry about I/O. Except for logs, of course. Those things are incredible disk space h

Heh, most systems could fit their entire working dataset in 1TB and never worry about I/O. Except for logs, of course. Those things are incredible disk space hogs. Also, serialising data to logs can surprisingly expensive. (Our ES logging cluster has tens to low hundreds of TB of storage, and it expires data after roughly 40 days.)