I recently upgraded my communication processors and ordered them with windows server 2003. I am using the same communications software and the latest driver for the Accelport 8r 920 PCI from the digi site. I have two Accelport 8r in each of my communication processors (3 servers).
I am experiencing a system ‘pause’ when I have the software running. This is the same software I ran with the original communication processors (actually have also tried an upgraded version). The new machines do have dual processors (both with hyperthreading) so I thought maybe the latest driver would fix it, to no avail.
Any tips/directions I could use in resolving these ‘pauses’. I’m fairly confident they are related to using the digi accelport ports because with the software running/no communication polling going on the system does not experience any pauses. System loading (CPU usage) is usually below a couple percent.
I did not experience this problem under windows 2000/single processor system with otherwise the same configuration.
Could you please provide a better explanation of “pause”?
I suspect that your application is simply running faster on the new systems and then has to wait for additional serial data (which probably is not arriving any faster) to have some data to process.
Do you recall if in the older PCs you were using the driver installation defaults?
You may wish to try changing the Receive Event delay setting from the default of 10 to a lower value such as 1 or 2. This will slightly reduce the latency of receiving data but it will also slightly increase your CPU usage, which does ot appear to be an issue.
This is done through the device manager, multiport serial adapters, AccelePort Xr 920, properties, advanced, select a port, properties. This is a per port setting, so be sure to clone or copy the settings toall ports.
the application is IO Server software, I use it on other machines (test machines) without digi ports and have not had similar issues. so I’m kind of targeting the digi driver right now (note, the other machines only had xp, not windows 2003 server)
The pause is seen system wide, not just in the application. Start button/IE/heck even notepad just seems to hang for a moment. After a few moments processing resumes. I don’t see any spikes in cpu usage/etc.
I have tried many combinations of event delay settings, close port timeout, ignore tbusy, complete writes on txmit, enabling/disabling error checking (and yes, I copied the config to all the ports).
How frequently are you seeing the system slow down like this?
(Every minute, 10 seconds, ?)
If you use the task manager to monitor the networking and the memory usage during these times, what do you see during these slow downs?
If you can see unusual activity, check the processes and sort by CPU so the most used process is topmost. Which process appears to be using the most CPU cycles?
If you then sort by Mem Usage, are you seeing a particular process increase its memory usage?
If yes, what is doing so?
You should also check for any hardware resource sharing or conflicts with any resource assigned to the any Digi adapter. Do you have any shared of conflicting resources?
Comparing using an application on a single serial port versus using 16 ports is not very helpful. Try using just 1 Digi port with your application and see if you se the same slow down. If you do, does it last for the same length of time or does it last for 1/16 the time when compared to using all 16 ports simultaneously?
Well, after some trial/error the only thing I am seeing the CPU usage spike on one processor (not one all the time), kernel time follows it nearly identically.
No conflicts or anything reported.
I’d say this occurs maybe once or so a minute. I do not believe it happens on a regular interval.
no processes growing memory wise.
difficult to try just one port on my live scada system, but will see what I can do. any tips appreciated.
Sorry, I’m not quite sure what the next steps should be.
What is the approximate duration of these slow down periods?
Are you able to monitor all of the ports in use and see if it is one particular port which having an issue?
You could try running Port Monitor to capture a port’s activity if you could narrow it down to port or 2. Capturing the data from 16 ports makes the results very difficult to interpret.