How to make 1C save a resource?

Question

1C

Hello.
At the office, a rather powerful computer was launched, as a server for 1C.
Employees connect via RDP and open the required number of bases for them.
Everything would be fine, but there is a problem - each of the working databases sucks out one and a half gigabytes of RAM.
On average, two employees work, and each has 10 simultaneously open bases.
Perhaps there is a way to reduce gluttony, at least by 300-400 megabytes?
Perhaps there is a way to unload from memory those databases that hang in the background?
Please share. Preferably, by methods worked out in practice;)

Answer 1

I have the following schedule. 1C cheekbone. 6 bases. The number of simultaneous users is 22-25 people. Everyone usually works with one. Bases approximately on 4 - 4,5 GB. The configuration is very non-standard (I will not go into details). The approximate consumption of the rphost file is 2.3-2.5 GB. Each user - 500-700MB. SQL server - 12GB (as much allocated). Lots of full text searches.
Such low memory consumption (IMHO) was achieved as follows: as the 1C programmer said, they have a de facto standard for storing binary data in the database (stream), all requested data in memory for as long as possible. Hence the decision that the binary data (in my case, these are photos) were cached in memory for a very short period of time, I don’t remember exactly, but no more than 5 minutes. In no case were they stored in the database, but in a separate disk partition. And in base only references to them.
Strain 1C-programmer!!! If they themselves are not strong in this. It is also worth asking / demanding / ordering him to optimize some heavy queries. This can usually be done.
PS. Guys of the same class, nothing personal, it's simple you are sometimes a little lazy.

Answer 2

Unused memory still goes into the virtual machine.
Just buy more RAM - it's cheap.

Answer 3

Judging by the gluttony, do you have file bases? If file, go to the client-server version. The memory will be eaten by the server itself, but each client session will consume significantly less.