One of the big differences between traditional databases and NoSQL databases is that in many cases it is acceptable to loose "some" data in NoSQL database. What this means is that write semantics are looser. They do not always write the data. Specific types of data loss are acceptable based on the NoSQL database type.
Do you really need all of your data?
You are collecting"
Log files of http requests
Log files of http tracking data
Log files of network performance metrics
Phone numbers dialed in the US: Who, when, where, how long
Much of this data is collected, processed to extract useful information and then discarded after some time interval.
What does it matter if you loose some data?