Hi @Alexei Konoferchuk ,

I had a similar issue in Caché 2012 on a Windows enviroment. I my case I got a different error code, but the same error description:

System.Data.Odbc.OdbcException (0x80131937): ERROR [08S01] [Cache ODBC][State : 08S01][Native Code 461] 
[F:\Matrix\MatrixDiagnosis\MatrixServico.exe] 
Falhou a conexão via cconnect: 
TCP connect() failed - exception satisfied select(). 
Reason: (10061, 0x274d) No connection could be made because the target machine actively refused it. 

Note that the Native Code 461 is a returned exception from the Intersytems ODBC driver, but the error message is thrown by Windows TCP. That means that we have to understand why this code is different for your scenario, but this Windows message has some known reasons:

1. If Windows is up, but Caché service isn't;
2. You reached one of Windows TCP limit (check this link for more discussion about);
3. The CPU workload is too heavy and can't process the new TCP connection, resulting in a full TCP backlog (check this link for more);
4. There are network issues, like a misconfigured virtual IP or a duplicate IP address that could make the ODBC request reach a wrong server and reach the case number one.

So you will have to understand if you a reaching one of those scenarios. If you have a high CPU usage, maybe you are possibily entering in the case number 3. I would check the CPU queue length (using Perfmon) to confirm. If you there are just workload peaks, then a Windows TCP tunning and a backlog extension in Caché could help you. If this is a common case, then you will have to check your CPU resources.

In the fourth case, you would have others ODBC errors before get this error... 

Sorry about my bad english! Good luck!

Wow, thanks, thats a good tip!

Hello Eduardo,
Thanks for the tip! I had read some of the articles and they are all great, all very good. I'm sure I'll use the knowledge of the @murray articles to add to my support and scaling activities. There are articles that @murray explain how to compare two different server proccessors, but how to understand the server as a whole?

I would like to find a method of comparing two different infrastructure environments. My clients create different workloads profiles during their production, so compare those servers is very difficult for me. I would like to have a standardized and well-accepted formula for:
- understand the power of a new server before deploying my system;
- understand the server / envoriment response after a change, such as a hardware upgrade or software parameterization.
Possible physical and logical bottlenecks can be evidenced even before testing on my application by the client or my team. I think that even though it does not contain the details of my system usage, that kind of tool can create parameters for future analysis. 
My question is whether TPC-E can be used as a solution for these cases. Is there a similar tool for the Caché world?