How to increase speed for initial call up of a PG2 Display
I have been asking several questions about the performance of the ABB System 800xA System in the past few months. Some examples:
Thanks a lot for your reactions, it has helped us a lot. At this moment most of our PG2 Graphics have a load time of <1 second. Most of our problems were related to using Aspect View Wrappers in Graphics, these were killing the performance of our displays. Besides that there were some network issues that have been solved.
There is only one thing left that I am trying to improve on our system. The load time for a Graphic when it is called up for the very first time on a client is stil between 5 to 10 seconds. Bottleneck for this is the CPU load of the AFWADServer.exe service on the Aspect Server. Since the AFWADServer.exe service is a Single Thread service one thing we can do is installing a CPU with a higher clock speed. But are there also any other options left to improve the speed for initial call up of a PG2 Display?
Whenever you call a display from the operator workplace first time offcourse it will take little while depending on the subsriptions used, since it fetches from the directory later on the cache will get created in the local Machine so later on calls are faster.
You try improve the higher physical RAM and graphics card, I never tried earlier
I would guess that the client computer's processor speed (GHz) is a major factor here.
At launch, workplace will pre cache all NLS Resource Data aspects (I once witnessed what putting hundreds of large JPG images in there could do...)
PG2 is based on WPF which in turn is a .NET 3.5 thing. The first PG2 aspect called will result in a "pyramid" of WPF and .NET initializations in the workplace. Again GHz is preferred.
I would not use a SSD in a server for application data storage (it risk wearing out before a mechanical disk since number of writes vs MTBF is typically less for a SSD than a mechanical).
But in a less sensitive node such as a client, a SSD will give a well needed boost.
If still in use, I recommend jettison the Default Aspect Key from the Aspect Precedence List. If/when used, it will use up more RAM than if only aspect categories where iterated in the hunt for what to draw on left click.
I don't think a SSD drive in the server will help a lot in the performance. When I have a look at the kernel times (Windows task manager -> view -> show kernel times) these kernel times are only a fraction of the CPU load.
@Stefan: interesting stuff about the pre cache of NLS Rescource Data, that's new for me. But I don't mean the launch of the workplace itselves but the navigation between Graphic Displays after a workplace launch. When I am navigating between displays after a workplace launch this takes a lot of time. A few minutes later, when I am navigating through the same displays then they are fast.
@ Hardha.D: the RAM on our clients is 8gb and we are using high performance Graphics Cards with 4 gb ram. So I don't think that is an issue.
So if I am correct, the initial load of a Graphics Display takes longer because it has to fetch the subscriptions from the server? After this the subscription paths are cached on the client so the client can connect to this subscriptions faster?