“Centralized Relational Databases” vs. “Decentralized Blockchain ” : Which one is really more scalable?

Question

There are many research papers on problem of scalability of blockchains:

Just as an example : An Overview of Blockchain Technology: Architecture, Consensus, and Future Trends

However, (usually by some companies which uses blockchain technology), it's mentioned that one of the advantages of blockchain against centralized relational databases, is scalability.

Just as an example : WePower White Paper.

Quote from this white paper:

Relational databases are adequate in many applications and situations. However, sustainability and scalability are limited with these databases. From the perspective of sector transformation in trading and digital infrastructure creation, transforming energy production to 100% renewable, the answer was to utilize a public blockchain.

Is it a correct proposition?

Answer 1

(Note: My answer is under the assumption that "Blockchain = Bitcoin Blockchain." This doesn't apply to file blockchains such as SiaCoin or Storj)

It depends on how you define "scalability" and "centralized".

Centralization

If you mean

centralized = on one large computer

Yes, the blockchain might be more scalable, as it can be sharded: Here's the huge Ethereum sharding FAQ.

centralized = many computers in a large company

Relational databases can also be sharded, thus scalability depends on the type of the data.

 Data type

Immutable data / data with version-tracking

Blockchain might or might not be a better option, depends on the implementation.

Mutable data

It's not a good idea to use an immutable blockchain to keep mutable data. You can just push the changes to the blockchain. On the contrary, you will have to scan the blockchain from start to end if you need to access one file, unless the current files are so small that you can keep a copy of it. (<- Which are bigger, the files or the version history of files? It shapes how you define scalability.)