It would be nice to have the option to select either starttime or endtime of the eventframe as the timestamp for the value that will be written a PI tag using the Outputs at close feature in Event Frames. Currently, it will only write the value with a timestamp of the endtime of the event frame.

Use case:

I have an event frame and I would like to see the average of an attribute value for that duration of that time frame on PI Vision. The expectation is to see a horizontal line indicating the average value throughout the duration of the event frame.

As of now, the average value is written to the underlying PI Point with the timestamp at the end time of the event frame using the Outputs at close feature in Event Frames.
This leads to a sloping trend when you try to visualize the resultant average values. i.e. the trend links the previous average value (calculated from the end of the previous event frame) to the latest event frame output average value.

With the requested option available and combined with the step parameter in the PI tag, we will be able to visualize a straight horizontal line for the average value of the tag for the duration of the event frame for that particular duration.

This is a followup to discussions with Roger Palmen and Rick Davin about their excellent writeups on the topic of EF Duration attributes; see Event Frame duration - easy way in AF2.6 or higher & Duration attribute on an Event Frame - Part 1 , respectively.
With the addition of the %Duration% substitution parameter and it's support in the Formula Data Reference, the following is suggested as a standard way to add a Duration attribute with values in a Time class UOM for newer versions of AF.
Create an EF attribute template with the following configuration:

Note 1: no parameters need to be defined; the substitution parameter is put directly into the Equation and can even be retrieved from the "helper" menus (red box and arrows).
Note 2: As mentioned in Rick's blog, the Formula DR supports unit conversions so we can change the result in seconds produced by %Duration% to any other time class UOM (from seconds to hours in the example above; blue boxes).
Note 3: A third method not shown here or the referenced articles, is use of EF Generation analysis outputs to add duration (see KB01884). Pro: this is a nice, all-in-one method if you also want to output the duration to a PI Pt attribute on the reference element. Cons: Because the output expressions are not evaluated until the EF closes, there are no intermediate Duration attribute values available for open EFs. This method seems more complex to configure.
For further details/discussion, please see the comments under Rick's blog.
Please vote/comment here to let everyone know what you think of this as a standard way to get EF duration into an EF attribute.
Thanks for your time!
Edit: It looks like some automated process to convert PI Square ideas to UserVoice entries removed the links & screenshots; added them back.

A few minor improvements to the search and filtering capabilities of the management tab could save time when doing backfills at scale - for example after transferring several elements from a development server.  If possible the following search capabilities could be very useful:
-Improve search by template: Currently when searching by template every analysis will be listed and in alphabetical order by analysis - not template name which makes it difficult to search. It could be helpful if it were possible to choose one template and every analysis associated with that template would be included.  At the very least, adjust the ordering of the drop-down so that it is in alphabetical order by template
-Improve Element search: Include or allow for search of child Elements as part of Element search

Originally posted on PI Square at https://pisquare.osisoft.com/ideas/29295-individual-collective-members-health-data-matches-with-other-nodes

We would like a way to know if a node in a collective is not receiving the same data as the other nodes. It's my understanding that all nodes in the collective receive very similar data. The PI collective node members should be able to talk with each other to know if one node is not getting similar data.
 
At times, a node receiving data from PItoPI or other methods, may not be getting data. The data could be buffering (a single path), hard drive space may have ran out on the PI server,  or a network issue may stop the individual PI server from receiving data.
 
Many times, we don't catch this if the other nodes of the collective did receive the data. Whether it backfills in or not, we would like to know the health of the collective and have this as an indicator a node is not healthy.
 
To add, Programs like the WebAPI or even ProcessBook don't know if data is not correct on a collective member. If there was a way to capture if a node is not healthy, it should filter into the other programs accessing the data, and steer "clients" away from using that node, as if the node were offline.
 
I have a hard time with a HA architecture when a node in collective may not have accurate up to date data, and a client connected to it doesn't know, therefore providing bad data to a client or other services polling the data.

It can be difficult to locate an analysis group in the current (2.9.5) UI for the Analysis Statistics in a large system. Currently the groups are organized in decreasing order of lag which is great to troubleshoot latency issues. However, if one wants to keep an eye on the performance of a particular group, they must scroll through the list to find the group which can be painful in a large system. This comes up frequently when trying to see if a group is has some skip counts.

One way of improving the UI could be to include the type of analysis along with the group name (Expression, Rollup or Event Frame).

Right now BadVal can only check one variable or Expression.  So if I am evaluating an expression with say 10 inputs, and I want the expression to output NoOutput() if any one of the 10 inputs are BadVal, then I have 2 options right now. 

1) copy the whole expression into the BadVal funtion and do "if badval(expression) then NoOutput() Else expression".

2) Do "if (BadVal(input1) or BadVal(input2) or ...) then NoOutput() else expression".

Ideally you would have a IFERROR function like in Excel.  So you would just type "IFERROR(expression,NoOutput())

Alternatively you could have BadVal allow multuple inputes.  So "If BadVal(input1,input2,input2,...) then NoOutput() else expression"

I'm sure one of these options is more resource efficient then the currently available options.