Storage with fast read (slow write), able to store large amount of records

Question

I'm searching for fast key-value\structured storage for lots of long strings. I'll update them seldom (and completely).

I (will) store some often-update data (as users ect.) in MongoDB; as run-time update small DB (for chat) I ('m going to) use Redis (it's so pity, AFAIK, there is no ability to store data not in memory).

Why won't I simply store it in MongoDB? Well, I'm optimization-maniac :), so, when I found out, that MySQL is faster than MongoDB in simple key-value reading (from some articles; also was said that MongoDB was designed for very fast write, not read) I was a bit disappointed.

So, what about fast for read storage that can store really lots of information.

Also, it's a C++ project.

PS: what about storing it as files? What if there will be really A Lot of records.

Answer 1

Have you tried benchmarking the sort of activities you're likely to perform on the large strings using MongoDB? I wouldn't write if off without performing some tests beforehand - better to try it out than to go off and implement your own storage strategy that results in a disconnect between your data sources.

What's the gist of your C++ project? Are you really going to need bleeding edge performance? Write now, optimize later is the best strategy IMHO - you can waste a lot of time writing 'optimal code' that has little or no bearing on the performance of the application as the end of the day.