Am I wrong in assuming that the responders are JRebel users? I'm asking this question because I wonder how representative is this sample compared to the industry in general (I know, this could be asked about any survey, really).
For instance, use of MySql with 32% seems too high to me, this, compared to what I've seen around. The main persistence solution of 1 out of 3 of my connections is not MySQL.
That's a great point. JRebel's the kind of tool I imagine a lot of middle-sized (not startup sized, not yet enterprise) codebases probably haven't evaluated, so I'd imagine it's tied to MySQL being a frequent choice for new apps and the quick choice for a one-off internal app in the enterprise.
While this is anecdotal, like you I've find the DBs most of the Java apps I encounter aren't MySQL, but a mix of Postgres, MSSQL and Oracle.
Unfortunately this is exactly the kind of anecdotal which just doesn't work. Popularity of tools and technologies varies greatly between geographies, industries and maturity. E.g. English-speaking world and German-speaking world value tools and techs in entirely different ways.
Also, note that MySQL was the only half-way trustworthy free database in the market. The popularity of PostgreSQL has been on the rise only since a few years. So if "free" (as in beer) is what you're after, then MySQL was the only choice for a very long time. Once you build upon MySQL (or any DBMS), you don't switch easily...
Given that RebelLabs is not mainly focused on JRebel users (it's more of a general ZeroTurnaround content marketing platform), the target audience is certainly JRebel-friendly. So yes, the results are biased towards that sort of audience.
Hi, Oliver here (the report's author). Actually, we specifically stayed away from contacting JRebel users simply to avoid any additional bias. Our charity approach (each completed survey donated to a charity that brings video games to kids in hospitals) that we didn't need to promote it as heavily to our own contact. We spread the word successfully through contacting RebelLabs subscribers, VirtualJUG and London Java Community (LJC) members and individuals in the industry. That said, there are certainly some JRebel users in there, although we didn't ask specifically.
It was hard to install for shared hosting, so most aspiring developers became familiar with MySQL first, with resulting network effects.
The client was a bit harder to use (\x commands are harder to memorize than mysql's "SHOW TABLE" and felt sluggish in a local environment (something that's no longer the case AIUI)).
Replication was a third party addon at the time, and had a reputation for being hairy, though I don't have direct experience.
I wasn't using it, but it had a reputation for being slow and for the replication being considerably more broken than MySQL's -- apparently both of those have come a long way since.
For instance, use of MySql with 32% seems too high to me, this, compared to what I've seen around. The main persistence solution of 1 out of 3 of my connections is not MySQL.