Why does Tor use multiple directory nodes instead of something like multiple cloud databases? Because then DDoS attacks couldn't take place on the directory nodes, as there aren't any, and cloud services like Google Cloud support concurrent requests onto cloud databases; wouldn't this be easier than having directory nodes that have to always stay online?
First of all, read about all of the things that directory servers do:
See Sections 6.3 and 7 of the Tor design document.
I think the most cogent things are:
- Directory servers aren't just any old servers that are randomly picked, they are trusted servers.
- One of the possible attacks that concern the Tor developers is, an attach that will "Trick the directory servers into listing a hostile [Onion Router]. Our threat model explicitly assumes directory server operators will be able to filter out most hostile ORs."
Your question assumes that the Tor developers could possibly trust a service like Google Cloud to be a good and trustworthy holder of the kind of data that could bring down the entire network. I don't think anyone who cares about the Tor network would outsource that kind of responsibility. When you also read about the other things that the directory services do, you would still need those services anyway.
Well, if something Truly decentralized - like a PoW blockchain registry - is used, then current architecture of the Tor network is not a bit worse in comparison with a cloud databases: we must either eliminate a centralization totally, or accept a trusted portion of it. Cloud databases - to be able to be successfully queried from any place and without any pre-seeded consensus will be the same centralized case