[Progress Communities] [Progress OpenEdge ABL] Forum Post: RE: Issues with PAS and network connectivity

Status
Not open for further replies.
D

dbeavon

Guest
>> there is no need to worry about thread starvation and it just blocks So why can't PASOE "just block"? If I read the explanation on that side of things, it seems you are saying that the wait on a forked process is a high-CPU operation? That doesn't make much sense. Doesn't the operating system have some kind of notification to wake up the parent process (parent thread) when a child process ends? As a side, I do appreciate the fact that these things aren't normally done for a malicious reason. The problem from my perspective is that people may use performance to say that Windows is better than Linux or vice-versa and, in the end, the difference may be as simple as the fact that one of them has a sleep statement that the other doesn't. When we were considering where to move our OE databases (currently on HP-UX) there were people (even some Progress experts) who said Linux is "10%" superior where performance is concerned compared to Windows. When you ask what the 10% overhead consists of - network, disk I/O, or whatever - and they can't/won't tell you for certain.... In the end, it might be that a sleep statement has been introduced in one and not the other (for whatever good or bad reason). Another problem with these types of cases is how hard it is to escalate performance problems when working with tech support (for any software vendor). These cases can be hard to repro consistently and, even when you do, they are much lower priority than anything else that the vendor may be working on. Memory leaks are the same way. I don't think Progress tech support wants to investigate PASOE memory leaks unless they are *extremely* rapid leaks (eg. 100's of MB per _mproapsv per day).

Continue reading...
 
Status
Not open for further replies.
Top