Means, at first I will answer not on business, then on business:
1) basis slightly cleaned. As far as possible. Its size reached before to 180 + Gb. So the cache in 200 Gb is quite justified, considering that the server will work further years five and the basis will grow.
2) at the height of the working day of 300 connections from which to steam of tens with analytical samplings (the big requests with groupings and sortings). Simple switch-off of a cache gives in rate the directory some tens sortings with the sizes from 1 Gb and above. So 64 Gb under rate a cache are a severe reality, instead of imaginations.
3) if you track tendencies the SAP passed to basis of SAP HANNA, which "in-memory, column-oriented, relational database management system". Key here "in-memory". On responses, on practical examples at them acceleration in 100-1000 times went. So great volumes of storage it already the reality, and will be tomorrow daily occurrence.
Now on business:
1) left for the night gbak too it was completed by an error. A screen in an attachment.
2) but, launched on other server gbak it was completed successfully. The total size 126 Gb.
In what a difference:
The server 188.8.131.52917
Keys-v-o at recovery. It was recovered long, one and a half days.
On this server of 32 Gb of the RAM
Amount of buffers 2000, remaining parameters firebird.conf on-default.
Now I think to try to return all configuration settings Firebird on-default and to run recovery on the big server with the buffer of 2000 pages.
If transits, then business in great values of parameters. If is not present, then in something the friend.