GreenSense infrastructure upgrades

[John CC]

The current GreenSense infrastructure is fairly powerful. I'm yet to see any open source software/hardware projects with this level of infrastructure and automated testing.

But I've built it on a budget and so much of it is slower than I'd like it to be, and it's limited. Automated tests could be a lot faster with more powerful computers. And there are a bunch more automated tests I want to implement.

The 3 garden systems involved in automated install tests all run on Orange Pi Zeros. These work fine for a live system but installation is slow installing to the SD card.

The hardware and software test servers are both installed on a single Odroid XU3. While this is quite a fast single-board-computer, tests speeds can still be improved with even faster boards with a SSD drives instead of an eMMC card.

So I've ordered a whole bunch of new single board computers and SSDs to make some major upgrades.

Here's the spreadsheet showing my infrastructure upgrade plans, and showing which hardware I've ordered:

https://docs.google.com/spreadsheets/d/1sipl5Pg57jkERKsEqMl6X8-F9PNRC2762IJf0HzigDM/edit?usp=sharing

I've also ordered a bunch of other bits and pieces including:

- a $200 1000w AC/DC power supply which will power all these systems until I get my solar system up and running. Then it will become the backup power supply which will turn on if the batteries get too low, to charge the batteries and keep all the systems running.

- some other small bits and pieces like voltage regulators, DC connectors, etc.

Soon the automated hardware test server and software test server won't be running on a single Odroid XU3 with an eMMC module, I'll have a dedicated Odroid HC1 with SSD for the hardware test server and another dedicated Odroid HC1 with SSD for the software test server.

This should mean I can start having some tests run in parallel instead of queuing and having to wait for one test to complete before the next test starts.

It should also mean I can add entirely new suites of tests such as sanity tests to double check live systems are working, and system level install tests to various types of RPi to ensure they're all compatible.

I also ordered another Odroid HC1 and 1TB SSD to use as a web server for hosting the GreenSense website, and a bunch of services like riot.im, a wiki, and anything else we need.

One day these services will likely be moved to a cloud hosting provider, but for now my broadband is fast enough, and hosting it myself keeps running costs low. A cloud hosting provider can be considered once GreenSense has a steady stream of income.

Keeping running costs low means the project can never run out of money and go bankrupt. I'm already paying for my broadband, the only additional running costs will be a bit more electricity. And as my solar system gets implemented I'll be getting a fair bit of electricity for free.

My two Odroid XU3 boards will replace the OPi Zero boards in the dev staging and master live garden systems. This will speed up the automated install tests and should cut minutes off the test times, as well as increasing reliability of the test systems.

I was going to order an XU4 for the dev staging system so my XU3s could be used for the master and lts systems, but HardKernel is out of stock. So for now that'll have to wait.

I haven't yet ordered all the RPis on the list because I'm running low on funds again. I'll order them when I have the funds. I need to prioritize the remaining funds I have to complete the installation of the hardware I already have (including my solar system), and the hardware I've ordered.

These upgrades should not only improve reliability, speed, etc. of the infrastructure it should allow us to add a lot more devices, more tests, and a lot more functionality that currently isn't supported by the infrastructure I have at the moment. So it really will be a step up in terms of bringing GreenSense from a small open source project, closer to a commercial level project.

The additional testing I can add will help ensure the systems are rock solid. And the improved test speeds mean any time a bug is found I can fix it far more quickly than on the existing systems.

The OPi Zero boards currently in use will likely end up becoming garden computers for more live systems that I deploy in the gardens around here. They work fine as garden computers once the system is installed, they're just annoyingly slow to repeatedly run install tests on primarily because of the SD cards, which is why they're being swapped out for Odroid boards with eMMC cards.

If anyone else wants to help invest in infrastructure upgrades and/or live demo systems there's a lot more I have planned.

I'll post more info as parts arrive and I start installing everything.

Cheers,

John