Percentage of elapsed time that all the disks were busy servicing read and write requests on the SQL Server computer.
Have added in extra caveat that there must be high latency as high usage but low latency is not neccessarily an issue depending on what activity is happening. Such as Transaction Log backups etc.