Doug, welcome to the best IoT Pi Linux OS forum
On the B+'s mine max the CPU to <6% when updating the webpage.
This is once a minute as that is my data read rate.
The rest of the time <2%, not an accurate measure of CPU usage just an indicator.
Rev 1.2 of Pi2 has the same BCM2837 but under clocked to 900MHz.
Under clocking a Pi 3 should help with heat.
But when you need Raspbian desktop and chromium every bit counts.
Check the Browser post for my venture into TC/X11 desktop with piCore.
piCore/X11 is much faster than Raspbian on a 3, so I am seeing if it usable as a development box.
Midori when in repo should help, Dillo3 is ok for text pages.
Bela has put my favorite editor Geany in now
A lot of stuff gets installed for X11, don't need it for remote if you can put up with the retro look of MC.
MC is a bit of a pain for developing but ok for fixing things, mostly it is copy and modify now.
HTML5 can be developed on PC/Linux/Pi.
Filezilla is great for copying the files to the remote box.
Most of the setup is done in the /opt/bootsync file.
Scripts are run with cron and watch.
Cron for slow scheduled stuff, watch for scanning GPIO pins for button presses.
Micropython for toggling GPIO for buzzers/sound.
When I first started with piCore I was behind a firewall and installing tcz's was nearly impossible.
So I started seeing what I could do with just the basic install. A lot more than I expected it turned out.
Even did the web page plotting with shell generated inline SVG
But life is too short to reinvent auto scaling graphs, so i used a JS library.
http://dygraphs.com/It plots csv files better than Google charts too.
No data base needed just daily csv files.
Old Linux stuff like Shell scripting with Awk, Sed works fast enough.
Busybox has lots of useful stuff in it.
I use shell script and netcat (nc) to send emails for alarms and status.
Because piCore runs from ram, reading/writing ram files is fast.
I can get one bit of script to write text into a file subject.txt, the message into message.txt etc.
Then another shell script reads the subject/message and inserts then into another preformatted string to a email file etc.
Basically I am using files as global variables.
As for a central storage, not convinced yet.
The piCore OS does not take up much room on 8GB micro SD's,
I use the rest to make a third partition /data/logs and symlink it to home/tc/www/logs so busybox-httpd webserver can find the data.
2 years of data logging still have >90% of 7+GB free, have not replaced sd card yet.
Have one old model A that's 4 years old still running really old Raspbian OS/python code.
Because it has no network it never writes and does not log either.
This means I have local storage and data plotting on each Pi.
Later on I can always copy files to a central file server, still need to figure this out.
I think the industry is still traditional and wants the data in one place.
I have lots of wasted cpu cycles, so i am looking at deep learning and data/trend prediction running on each Pi.
Example - I monitor a bunch of fridges, I can check and analyze the time between compressor cycles.
This is obvious from the temperature plots, i can tell when the door opens and if the room aircon is working etc.
I should also be able to tell if the cycle times are getting shorter, this would indicate coolant leak.
I can have one Pi monitor 4 fridges, more if I multiplex.
Predictive maintenance.
Sensors are getting smarter but once they have Pi's in them it becomes game changing.
Remote desktop is not so important as remote data viewing, which HTML5 stuff does fine.
Next step WebGL for plotting trends over day/months/years?
But that requires someone to look at the data and make a judgment - 1, 5, 20, 500 fridges
Better if we get the sensor to make it's own decision and send a message, " hey, I think I am about to fail" or "someone left the door open", "the cleaners unplugged me so they could vacuum".
Now if you want to do real time control with Pi's with timing loops in1/100sec, nearly possible but the Linux OS gets in the way.
Which is why I spend lots of time relearning Pascal for Ultibo.
Or just off load it to a Arduino or my preferred chips - PSoC's.
PSoC's are useful as they are programmed via I2C, the same I2C I use to talk to them by the Pi.
The Arduino IDE runs on Raspbian and even does the Adafruit LoRa Feathers which are great for long range RF sensing.