D
dbeavon
Guest
>>If I increase the amount of agents and limit the agents to 2 sessions each i start seeing a noticeable improvement on the CPU load. This forces requests to be serviced by multiple agents rather than just one. So this tells me that the problem is not related to the virtual machine. It has to do with the fact that your ABL sessions are sharing living space with one another within the _mproapsv process. They are conflicting with each other and they should not be (except for certain rare scenarios that are well-known, eg for xml and the clr bridge). I would definitely review your 4GLTrace outputs and isolate the specific parts of code that get unreasonably slow. You might even try some artificial tests that (1) eliminates the database access altogether, or (2) switch from shared memory to client-server connectivity. I suspect you have some code or component that makes a bad assumption, given the nature of PASOE (eg. the ABL code might be assuming that it has an entire process/ID/resource/static all to itself).
Continue reading...
Continue reading...