Huge Firebird 5 impact on disk io

[resign1648357-ops]

Many simple select statements cause high write load in my drive.
It's a continuous load of a few MBps.

[mrotteveel]

That looks to me like a write to the header page each time a transaction is started. You have a 32KiB page size by the looks of it. This is entirely expected and normal if you (or a driver in auto-commit mode) starts a new transaction for each select, and executes those selects near continuously.

[hvlad]

And why do you think it is not correct, or problem, or ... what ?

Note, you use biggest possible database page size, also, you disabled filesystem caching - no wonder it creates some load on IO.

PS usage of read-only transactions when possible will lower header page writes.

[resign1648357-ops]

Thanks for prompt reply and advice.
I backed up and restored my db with 8 kB page size.
My first thought was just the lifespan of the SSDs(TBW).
For my mostly read queries I wrongly assumed that I can just tune only one param DefaultDbCachePages(1st layer of cache?) and forget.
Yup I didn't enable filesystem caching(2nd layer of cache?) since it was disabled by default.
Is 2nd layer of cache a thing?

[mrotteveel]

PS usage of read-only transactions when possible will lower header page writes.

I thought that was no longer the case with ReadConsistency = 1

[hvlad]

PS usage of read-only transactions when possible will lower header page writes.

I thought that was no longer the case with ReadConsistency = 1

I speak about #5706 (CORE-5434)