> 1) We have sweep set to 0. We do monitor the transaction gab and if
> needed, we interact. Any kind of automatic sweep under high load will kill
> the server :)

1- Do you run a manual sweep on a regular basis?


2- What is the CPU % like when Firebird "slows down"?


3- Have you tried using CrystalDiskMark to measure your disk IOPs?

My local workstation SSD results are:

-----------------------------------------------------------------------
CrystalDiskMark 5.0.2 x64 (C) 2007-2015 hiyohiyo
-----------------------------------------------------------------------
Sequential Read (Q= 2,T= 4) : 564.293 MB/s
Sequential Write (Q= 2,T= 4) : 535.874 MB/s
Random Read 4KiB (Q=256,T= 1) : 227.652 MB/s [ 55579.1 IOPS]
Random Write 4KiB (Q=256,T= 1) : 195.258 MB/s [ 47670.4 IOPS]
Sequential Read (T= 1) : 545.480 MB/s
Sequential Write (T= 1) : 520.939 MB/s
Random Read 4KiB (Q= 1,T= 1) : 36.577 MB/s [ 8929.9 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 90.805 MB/s [ 22169.2 IOPS]


3- Have you tried to analyze/determine the long running SQLs being executed, to determine if wrong/no indexes are being used by queries or other optimizations that could be made?


Sean