You can use the Advanced button on the Analysis tab to set how the output timestamp is written. You can set the timestamp to be the Trigger Time (rather than the execution time so if the event reaches the server 1-2 seconds later due to network latency it will use the trigger time from the incoming tags for the timestamp).
10 calculations/sec would be a concern if you were doing more expensive calculations (i.e. moving averages over an hour for a quickly updating tag), but for something relatively simple like adding or subtracting two numbers I suspect you'll be fine and they won't lag behind. One more caveat on this, if you're planning on doing this for 10k pairs of tags all updating with a 100ms frequency that's a bit of a different story and you may run into issues. Analysis performance is a little bit played by ear unfortunately so I can't give you hard cutoff numbers, but from what I understand what you're doing now seems reasonable.
Thanks for your response. You have answered some of my questions.
The following questions are:
(1) Is it possible to determine both of tag_src1 and tag_src2 have had values at time1 (supposed that they both have at identical time up to millisecond) so that we can perform operation like (tag_src1 + tag_src2) ? Since if we just use TagVal(tag1, time1) + TagVal(tag2, time1) according to tag1's trigger time, I'm not sure TagVal(tag2, time1) is available at time1 yet.
(2) As you said , there is no cutoff numbers for high frequency calculation. Do you recommend how we see the loading is still affordable and would not affect other PI functionality? e.g. just look at CPU / Memory usage?
You could do something along the lines of the following to check if the tags have the exact same timestamp:
IF ( PrevEvent(tag_src1, '*') = PrevEvent(tag_src2, '*') )
Yea, as a first litmus test, if you have the option to run these in a dev environment and see how they perform do that. If you need to run them in a production environment, we can always use the Analysis Service statistics to check whether the calculations are lagging behind (although you may want a Tech Support engineer to assist you with looking at that).