4 ms·
As usual, the confusion and religious comments are numerous. There is no such thing as "nosql". There are different types of databases, with traditional relatio
by manigandham 9y ago
As usual, the confusion and religious comments are numerous. There is no such thing as "nosql". There are different types of databases, with traditional relational being useful for 95% of scenarios (especially on increasingly fast servers with decent replication features) while the rest of the time something more specific is needed.
SQL is just an interface, obviously common to relational databases but can be applied to any datastore. Spark/Drill/Presto/Dremio/etc can give you SQL over any data, even just files in a folder somewhere, so let's clear up this notion between actual database technology and the access path.
Document stores are definitely useful. MongoDB is one of the better ones today although it had a rocky start. RethinkDB was an interesting experiment but never matured, RavenDB is a solid contender, Couchbase has proven itself, Riak might stick around, and there are dozens of others.
There is a place for everything and MongoDB is being used by plenty of companies to great extent. It might not always be the right choice but when it is, it works incredibly well. Good luck to the team, I'm glad to see the success in both the product and the company.
- philliphaydon 9y agoMarten > RavenDB!
- manigandham 9y agoIn some cases, but ravendb 4.0 is natively clustered and has more querying capabilities and performance when using all of the document store functionality.
- lstyls 9y ago> There is no such thing as nosql This is not at all correct. SQL has well defined semantics that are standardized around the relational model and ACID guarantees. A nosql datastore is one that intentionally makes tradeoffs that force it to deviate from that model. The name makes sense if you understand that SQL was the first broadly adopted language that targeted the relational model. I'm not very familiar with the systems you mention other than Presto. But if they do not provide relational guarantees then even if they have SQL-like syntax the semantics are sufficiently different for them to not be implementing a true SQL. Hence, nosql.
- manigandham 9y agoSQL = structured query language. That's all it is, a language to access and manipulate data and data structures. It comes in several variations, standards and dialects. Any system can implement (a form of) SQL as an interface, just as any system can also implement ACID guarantees around data access, but these are completely separate concepts. ACID isn't even mentioned on the SQL wiki page. [1] Relational databases have a history of offering both in a single package, but there is no "true" SQL. There are also plenty of non-relational data stores that offer SQL and/or ACID guarantees so this limited generalization isn't accurate or useful. "NoSQL" has no meaning other than originally describing systems that did not have any SQL access at all, usually due to different storage layers, distributed architectures, custom access protocols, and general immaturity around usage. A decade later, all of these systems have evolved and there's both convergence and specialization everywhere. It would be far better to just talk about the actual type of database, and the interfaces and guarantees it provides, rather than marketing jargon like nosql. 1. https://en.wikipedia.org/wiki/SQL https://en.wikipedia.org/wiki/SQL