Performance testing 1C

Question

linux

I want to measure the performance of 1С82 in several configurations

WinServer2012+SQLServer2012
WinServer2012+PostgreSQL
Linux+PostgreSQL

+ everything is the same in KVM, XEN, VirtualBox virtual machines. There is either a Gilev

test or take a working database (10GB), rerun all documents for 2012, measure time. Any other ideas how to test?

Answer 1

Gilev's pure synthetic test is not always adequate and does not measure multithreading. It is the same with re-hosting - 1 rphost will be loaded and, accordingly, one core.
When we conducted tests, we wrote special processing that performed typical operations performed by us (creating documents, generating reports, changing properties in directories, etc.) for several thousand operations-elements and measured the execution time. Plus, run from several computers (according to the number of rphost processes on the 1C server or more).
On the plus side, you need to take into account that different configurations require different things - some need a fast disk subsystem (trading in which there is a constant selection from directories), others have a processor (the same accounting department) as the main bottleneck.
I also recommend testing different configurations with a network and a disk subsystem, for example a) on the same database server + 1C server + storage, b) everything on different servers, c) 1C server + database server + storage, and so on.
PS: 1C has some fun with hyper-trading - the results can change several times when you turn it on and off, keep this in mind.
PS2: If the results are not under the NDA, then I would like to see them.

Answer 2

There is also a question about testing when the 1s server and the DBMS are on different machines (including on different virtual machines of the same host) and the impact of different “physical” networks (including the presence of routers on the way) on this very performance.