Aurora vs. MemSQL for low bandwidth but low latency

Question

We are changing the current MySQL user database. We have a website + mobile app for which US users request our database. the corresponding data is contained in three tables, and a join request for three tables is needed to send relevant results to users.

The result sent back to users is small (<6kb). If our goal is to switch to low latency, and throughput is low priority, which of the following two databases will work best:

MemSQL or AWS Aurora?

both of them have the same starting hardware costs (~ $ 0.28 / hour). We consider only these two databases at this stage, so that we can continue our "MySQL" knowledge of our own.

I like that I can transfer the DB headache to Aurora. But surely MemSQL's read / write ability is a solution with less latency?

Answer 1

Nothing beats memory for speed, and MemSQL was created for that. It stores tables (in rowstore mode) in memory and uses a custom query engine to cache queries in an intermediate language so that it can execute as quickly as possible. Aurora is more like a classic disk-based MySQL instance, but with more infrastructure and optimization changes to make the most of Amazon services.

Before making a decision, you need to find out what “low latency” means - is it for a few seconds or milliseconds?

MemSQL will be faster and most likely in milliseconds depending on your query. Aurora will be slower, but it may deliver a second second, again depending on your request and resources allocated and how the data is structured.

Without any details, the answer is to judge what your performance tolerance is and then experiment.