They're for the most part on my PC inside lightroom.
Then all that gets backed up locally to a synology NAS.
Also backed up via Backblaze.
I feel comfortable with this system, but it certainly isn't ideal for sharing and etc.
For all the posters saying Google Photos: make sure you have a backup of your originals. Google strips most of the metadata from your images when you pull down a Takeout or use the photos API, including GPS.
You can copy your original images and videos directly to a home computer/server/desktop/raspberry pi via either SyncThing or Resilio Sync.
If you want to self-host your library, there are a bunch of options, but (being the author) I'm partial to PhotoStructure: https://photostructure.com/faq/why-photostructure/
(I've had several beta users tell me that they've "tried them all" and picked PhotoStructure, but there's some survivorship bias there!)
I use Google Photos and I hope Google will not cancel it anytime soon, or ban my Google account for no reason.
I would recommend backing up with Google Photos and Backblaze.
For a local storage, I would invest in a hard drive. This way, you have two cloud copies for every local copy.
Family pictures are really really important, you will look back on these 10 years later and smile when you come across these again.
i have a copy on dropbox, a backup of that folder on onedrive with arq backup. For easy viewing I also upload them to smugmug and make sure they are tagged correctly. It works mostly in that we can find photo's most of the time and have not yet lost a single one.
iCloud Photos
Pasting my own answer from [0]:
---
For myself and my family, I wrote my own tool [1] that runs everyday on an "input folder". A quick google on "github photo organizer" shows a lot of others having done the same :)
It organizes all traversed photos by date (extracted from exif or from filename), and puts them in a "failed" folder if it can't parse the date.
If any photos get the same name, they are either deduped because they are exact duplicates, or are marked as conflicts (e.g. A.jpg and A_conflict1.jpg) if they are different.
Last time I used it for a large input it took 3h for 200GB, though I suspect network latency was the main bottleneck.
It's around 300 lines of python - verify the code for yourself if you want to use it! You probably also need to fork it if you don't intend to run it on a Synology NAS.
However, as I mentioned last time I pitched this, elodie [2] might be more suitable for others than my little hack. Haven't used it though!
1: (https://github.com/johan-andersson01/photo_organizer
2: https://github.com/jmathai/elodie
---
0: https://news.ycombinator.com/item?id=24019612