I am new to Server, and have some questions related to performance.
I have an analytics app that processes millions of records. These have long tables with 50-150 variables each.
The data are first transformed through many iterations resulting in the final tables.
The final tables are not joined ( no need with executeSQL() ).
Only 10-30 variables may be needed for any given "report" on the client side, generally.
Hence, I plan to use executeSQL() statements on the server side to build virtual tables that are passed to and joined (where necessary) on the the client side. No new data are generated that must be written back to Server.
I have a couple of questions:
1. Does FMP row size matter when processing data on the FMP Server itself?
2. Is there any reason my virtual table approach may be problematic, in principle?
3. How does Server process data differently from FMPA client?
4. What is the performance hit associated with encryption at rest?
5. Would the approach described above work also for iOS clients??
Question #3 above relates to my pre-use data transformation process.
This typically takes 15-hours, though there is a great deal of optimization to do to the routines.
In theory, many of the processes are not interdependent and could be separated and processed in parallel.
Might I use Server as transformation engine as well ... perhaps on different hardware. The transformations are done once each month. How might this be done ... "virtual" user accounts??
On the client side, the processes must be done sequentially.
We are thinking of using AWS on the server side.
Any input would be helpful.