How to configure IIS and SQL to reduce RAM consumption?

Question

SharePoint

Hello. There is a Farm Sharepoint 2016 consisting of 4 servers:
1) DB1 - WM, 8 cores at 2.3 GHz, 14 GB RAM, Windows Server 2016. SQL 2017 is installed;
2) SHPT1 - WM, 8 cores at 2.3 GHz, 16 GB RAM, Windows Server 2016. Sharepoint 2016 installed with roles - foreground and distributed cache;
3) SHPT2 - WM, 8 cores at 2.3 GHz, 16 GB RAM, Windows Server 2016. Sharepoint 2016 installed with roles - Application server and search;
4) OWA1 - WM, 8 cores at 2.3 GHz, 12 GB RAM, Windows Server 2016. Installed Office Online Service.
On DB1, the sqlservr.exe process eats up to 92% of RAM.
On SHPT1 w3wp.exe processes eat up to 80% of RAM.
On SHPT2, noderunner.exe processes eat up to 80% of RAM.
On OWA1, w3wp.exe processes eat up to 80% of RAM.
At the moment when RAM consumption becomes more than 70% - the farm starts to slow down a lot, only rebooting all servers saves.
I am sure that RAM consumption can somehow die, but I could not find a normal manual for fine tuning. Maybe someone can suggest something?

Answer 1

Server Properties (Memory Page) .

Answer 2

Look in SQL at you max memory is limited? If everything is fine there, look at indexing, especially content with RBS, interaction with files slows down a bit. insignificantly, internal processes for file transfer, archiving, delivery and related processes work.
You watched a profiler what happens at loading. There is also such a story, as soon as SQL starts to eat up, it eats everything it needs, and then this peak does not let go (especially with processors) for possible jumps.
The most interesting thing lies in the nodes of the search engine, there are 5 of them more precisely ( noderunner )
Check how often Full Crawl is done. If once an hour, then remove it once a day, you have a weak infrastructure for its operation (It works fine despite the fact that each of their nodes is on a separate server + some services are clustered). If you removed all the nodes and the entire search pipeline, albeit on a separate server, but that’s all, they take a bite out of 2 GB of memory in standby mode, eat the processor and load the servers + if you have a frequent Full crawl, they will load the database with eternal inserts, this is what you should have found out when profiling the SQL and finding out what caused it.
If, after all, you figured it out and are sure that these are SQL and eternal inserts and loads from the search side, then configure the search more carefully through PowerShell ONLY!