time table

No timetable for disk writes, it's not timed, it is just a queue, but there are many factors for a particular key/data-item including:

Disk I/O Speed [EBS is notoriously slow at times]
Rate of Document Mutation (when you modify documents it doesn't write every single change, only the most recent based on an algorithm and it's place in the write queue)
Rate of Data Influx
Cluster Size and Distribution (If you use fewer large or huge machines vs many smaller, i.e. Vertical Heavy vs Horizontal Heavy Clustering)
and probably a few more factors

What's interesting is that because of our architecture, when Amazon EBS had a cascading failure, many other NoSQL solutions went completely down, but Couchbase was able to continue to perform, with only the subset of data that was ejected from memory being unavailable. And when EBS came back up writes resumed without trouble.