What is the configuration of an ideal 1C server in terms of performance?

Question

1C

There is a server 1C and MS SQL, two in one and a certain base on it. 1C with MS SQL communicate through Shared Memory. About 200 people at peak work with the base. Around 20 people at night. Base about 200GB, logs about 5GB. There is a task to provide to all to it the maximum productivity. If there are no plugs on the processor and RAM, on monitoring, then there are questions about the performance of the disk subsystem.
Can you please tell me how to place the components of this whole bundle in order to achieve maximum performance in a vacuum?
Let's say:
Tempdb MS SQL - transfer to a virtual disk in RAM
Transfer the database logs to PCI Express SSD or also to RAM
Transfer the database itself to SAS RAID10 disks or something like that.

Answer 1

Shared works well for up to fifty users. There was a modest mention like in ITS. Simple analysis of resource competition: 1C application server: processor, memory; database server: processor, memory, disk; shared: memory. You can run into disk operations due to competition for memory. The bus of the north bridge of the processor is wide, but still not rubber. Therefore, it makes sense to switch to the classic three-tier system - move the application server to a separate hardware. You will immediately have about 30 Gilevsky parrots, but the behavior of the system will become more predictable and it will become easier to analyze and adjust the load.

Answer 2

200 users is not an indicator of anything. how many documents are generated per day? the average number of lines in the tabular part of these documents? what configuration? what is the reporting load?
1) tempDB - RAID0 SSD to take out (directly separate)
2) LOGS - another RAID0 SSD ('this is another array - you still need to put it on hold to mirror it on a SAS15K disk)
3) Data - RAID10 SAS15K + optain cache
4) folder temp + the 1C server cluster folder - move it to a separate SSD
and of course the disk controller must be appropriate for such IOs (maybe even two, otherwise it seems to me that pci-e 3.0 x16 will no longer be enough for such operations)