Why do most NoSQL DBMSs have “pointers”?

Question

What is the objective reason why most NoSQL storage solutions do not have any “pointers” for ultra-efficient joins such as pre-relational DBMSs?

I mean, I partially understand the theoretical reasons why there are bound pointers in classic RDBMSs (you need to update them and synchronize them twice for memory and disk, not for “disks” fast enough to be cured, for example, random access for some use cases, like modern SSDs, etc.).

But of the many NoSQL solutions, why so few of them understand that this model will be awesome (the exception that I know about would be OrientDB and Neo4j) for many practical cases, and not just for those that need graph traversals. I mean, when you need things like multi-join, you need Mongo Ping Pong Pong and execute N requests instead of one.

Is it possible to overlap a NoSQL db document with one of the graphical DBs, that such a function makes sense and simply provides all the practical possibilities of SQL connections for NoSQL solutions with a small number of additional ones and for most queries indexes are useless and take up much less space for huge data sets ?

(... and as a bonus, any NoSQL solution will be ready to use as a db graph, and tracing a path length of 100 nodes in a graph stored in Mongo will simply automatically work fast enough)

Answer 1

I believe the key issue is data locality and horizontal scalability. NoSQL's premise is that RBDMS redundant models, i.e. Those that require pooling lead to bottlenecks.

Twitter: , , , ( x x , [1-10M, - aplusk ]).

, , , . , , , / . - , (, , ).

, , , , , : - , > 95% .

, , ( / ): , , , , , node, .

, -, , , (!). , , .

~ 100 , ,

100 , 50 , . , , 5000 . , , .

, , .

, multi-joins, - Mongo N

MongoDB, . Multi-Join -, MongoDB. MongoDB. : Facebook TAO, , .