Hacker News Reader

Comment by jeffbee | Hacker News Reader

There are loads of big companies with huge, important databases in eventually-consistent nosql databases. In fact I’d say the bigger the company the more likely it becomes that they’ve got the software development chops to exploit that kind of database.

View on HN

I've observed the opposite: larger and higher-skill engineering departments are more likely to have the expertise and perspective to scale RDBMSes and use them in advanced ways.

It's smaller or lower-skill/fad-chasing orgs that end up doubling down on non relational systems for as the primary data store, lured in money-for-free promises of "no schema needed" and "scales automatically".

Nonrelational datastores have a place, of course, but they tend to be much more niche/purpose specific in functional engineering orgs in my experience--to the point that an engineering group's ideas around relational databases (what can they be used for? Should they be the default option for new use cases? How far can they scale?) are an effective proxy for that group's skill level.

We all have our own experiences! Mine have been that a mediocre dev team will assume the RDBMS does things that it doesn’t actually do. A good dev team will build their application around its own real requirements and then the database features they need are reduced to durability alone.

I've seen non-relational databases used for things like logs, streaming data, time series, caching. I'm not saying they don't have uses. You probably know that prior to the release of Oracle in the mid-80s all databases were non-relational. Some of the "NoSQL" ideas are actually old ideas, database models that got replaced by relational databases decades ago.

Lots of big companies use Redis, MongoDB, Cassandra, DynamoDB, etc. when those tools address a business requirement better than a relational database. I have not seen NoSQL databases replacing important relational databases, I have seen NoSQL databases used for specific business requirements.

Of course big companies do a lot of software development, so you're more likely to see a variety of languages and tools in a larger organization. That may or may not validate the tools some people in the organization choose. Many large companies still use COBOL.

Non-relational databases match one of the largest use cases that exist; large volume OLTP workloads. Non-relational databases typically can handle far more load on reads and writes than relational databases due to their intrinsic distributed structure, meaning that if you have any kind of transactional workload, for example, a frontend to a busy store (such as Amazon.com), non-relational databases are incredible. You can also back these frontend non-relational stores with relational ones but at much smaller scale in order to handle the OLAP requirement where relational databases shine, allowing for massively flexible queries when needing to pull data for reporting etc.

No disagreement from me. Relational doesn’t address every business requirement, nor do non-relational databases. They have enough overlap that you have to make an informed decision from experience.

The SQL vs. NoSQL debate got framed early on in terms of better and worse, old vs. new, which is unfortunate because it created ideological camps. Not every supposedly new thing is better than the old thing. Some NoSQL techniques predate relational databases, but no one younger than 50 has experience with what we did before Oracle and DB/2.

Choice of tools should match the requirements, not fads or anecdotes or personal preferences (or ignorance).

"I have not seen NoSQL databases replacing important relational databases, I have seen NoSQL databases used for specific business requirements."

Look harder :-) Every single NoSQL vendor has examples of substantial "replace' projects.

Every single NoSQL vendor pretends that there is no difference between what you can do and what you should do.

I’m sure you’re right. I haven’t seen it. I have seen projects use NoSQL databases and fail when a relational database was a better fit for the requirements. The fault isn’t in the tools, it’s in the decision-making process. It seems common now for teams to choose a “stack” before they define what they plan to build.

Reply on news.ycombinator.com