@ypercube: If the character set is UTF-8 and not Latin, it would be triple that – I don't think that's the case with Cyrillic alphabets. Actually, I was under impression that a single letter would take 2 bytes in UTF-8 for most non-Latin languages. Perhaps Eastern-Asian languages would take more bytes per character. But I'm not an expert in this area, that's why I'm asking here if you are certain about what you said in your answer.

@Andriy: You are right. But I think MySQL considers the worst case, at least when an index is created. Not sure what is done inside, how indexes are stored.

you can't have a VARCHAR(256) index for a example, because the limit is 767 bytes per index column.

You are probably right, though. I should rephrase to "it could be as worse as triple that"

@ypercube I'm sorry, I didn't realise you were talking specifically about calculating the index's storage size

So what I said may well be irrelevant

A pretty graph I made with Excel 2013, from a blog post coming this week:

I did a little post-Excel Photoshop work but that's mostly a stock chart

@AaronBertrand Nice. By CSV you mean a CSV string?