Below are some metrics with my new method. The internet connection used is not as quick as my school internet which my users use.
5 entities = 10,803 records total = 34 iterations of up to 500 records = about 22 real groups of 500
Elapsed time trials: 44s, 39s, 40s, 44s, 42s
- Average elapsed time: 41.8s
- Average number of records loaded per second: 258 records
- Average elapsed time per loop (iteration): 1.22s/loop
- Average elapsed time to cache 500 records: 1.9s/cache of 500
These tests were run on the same network as a previous test I ran. Below are some comparisons:
- There were 65% more records this time.
- There were 24 loops last time. The new test sample has 42% more iterations.
- The average time elaspsed this time is about the same as the fastest time last time (41s).
Therefore, the changes to PowerApps version 2.0.590 and this method have up to a combined 42-65% improvement in performance. Since I no longer need to write a column identifying which block of 500 a record belongs to, there are also improvements in writing. That represents about 50% less writing time since I previously wrote new records twice: once for the record with the main information, and second for calculating n as RoundUp(PrimaryId/500,0).
Although there is a performance improvement, I would still judge 40+ seconds of caching a long time for end-users even if it's a one time block at the start. If it is 40s right now, then with more usage of the app, the time will gradually increase to unsustainable levels. If Sum() or CountRows() could operate on any one of the entities in its entirety or if there were a one-click Cache() function, I could easily cut out at least 10s or more.
Microsoft Employee
@8bitclassroom
