@NosibleAI is data hungry.
It collects reference, market & alt data 24/7/365 from 00's of APIs and inserts it into a version controlled linked-entity graph of >120k listings and >100k funds & ETFs
Here's a short thread on cloud storage lessons learned
...
It collects reference, market & alt data 24/7/365 from 00's of APIs and inserts it into a version controlled linked-entity graph of >120k listings and >100k funds & ETFs

Here's a short thread on cloud storage lessons learned

Lesson #1 - AWS S3 and GCP GCS charge mafia prices 
Their price per TB per month is high, they charge you for every API request, and they rip your eyeballs out if you ever leave (egress)
If you have lots of smallish files, just kiss your budget goodbye

Their price per TB per month is high, they charge you for every API request, and they rip your eyeballs out if you ever leave (egress)

If you have lots of smallish files, just kiss your budget goodbye

Lesson #2 - blockchain is not just a buzzword. It is powering the future of cloud storage RIGHT NOW 
There are a few decentralized storage networks available but my favourite one has to be the Sia network @SiaTechHQ
PS. is it pronounced Sigh-a or See-a?

There are a few decentralized storage networks available but my favourite one has to be the Sia network @SiaTechHQ

PS. is it pronounced Sigh-a or See-a?
@SiaTechHQ allows anybody anywhere on Earth to rent their excess storage space to renters like me who would like use to it 
When you store a file it is encrypted, sharded, and stored on many devices on the network to ensure it is always available. It's magic

When you store a file it is encrypted, sharded, and stored on many devices on the network to ensure it is always available. It's magic

Lesson #3 - using Sia is 10x easier when you use @Filebase 
Filebase is an S3-compliant API that sits in front of both @SiaTechHQ and @storjproject. All of the benefits with none of the hassle
I trust and use @Filebase myself.

Filebase is an S3-compliant API that sits in front of both @SiaTechHQ and @storjproject. All of the benefits with none of the hassle

I trust and use @Filebase myself.
Lesson 4 - there are also some great centralized alternatives too!
Wasabi is great. It doesn't charge for API requests and is ~80% cheaper than AWS and charges no egress
The only "catch" is that they charge for a minimum of 1TB + 3 months. Which is a normal use case
Wasabi is great. It doesn't charge for API requests and is ~80% cheaper than AWS and charges no egress

The only "catch" is that they charge for a minimum of 1TB + 3 months. Which is a normal use case

Lesson 5 - you can also DIY
using @OVHcloud and @Minio.
If you need to store many 10's of TB then an even cheaper option is to self-host! OVH has cheap storage servers and MinIO is an open source S3 compliant API. Put those two together and you're winning
.

If you need to store many 10's of TB then an even cheaper option is to self-host! OVH has cheap storage servers and MinIO is an open source S3 compliant API. Put those two together and you're winning

After a lot of testing @NosibleAI I switched from GCS to using both @Filebase AND Wasabi. That gave me 2x live copies of the graph 
And guess what?
I ended up paying 75% LESS than for only 1x copy on GCS
and I have not experiences any issues.

And guess what?
I ended up paying 75% LESS than for only 1x copy on GCS


Bonus (negative) lesson - based on my experience I suggest you avoid @backblaze's B2 product 
Not everything that glitters is gold
you're better off using FileBase, Wasabi, and/or MinIO. And if you do use MinIO just be careful because you're your own support team

Not everything that glitters is gold

