Is mongo db suited for OLTP or Data warehouse operations ?

Question

Is mongo db suited for OLTP or DW operations?

What is the response time and effiency of mongodb in OLTP operations? Is it secure to use mongodb as a backend of a web application?

In mongodb itself we can do map reduce operation. Then what is the need for using hadoop integrator?

I need to know how strong is mongo db to use as a replacement for a BIG DW system ?

Answer 1

I'm using MongoDB for OLTP operations, currently I have 100+ operations/sec. and MongoDB can handle much more. In perfect case you can expect tens of thousands operations per second, but this number is hard to achieve in pratice.

Response time really depends on you replication/write preferences, because MongoDB gives you control over query execution(see CAP theorem). I don't know what you mean about effiency, I can say that insert operations is efficient enough(don't use updates for OLTP).

I have no experience with MongoDB security options, because all my web applications have full access to DB and I closed REST-API for public access.

Don't use MongoDB's MapReduce for large datasets, you have to trust me :). This is pain! I found Aggregation Framework suitable for large variety of operations with large datasets(Gb's of date). If it's not your case, try Hadoop's implementation on MapReduce, I don't have such experience, but always want to try.

As an option you may consider Hadoop's HDFS as main storage and something like messagepack as binary format. I heard about such solution.

Answer 2

Today, NoSQL databases are becoming more popular because the need for easily scalable, dynamic databases is growing, when you are dealing with cubes and huge volume of data. If you need to run at scale, RDBMS can not keep up with the ingest rate. RDBMS were not built for today's web/mobile/IOT applications.

To answer your question: for BI solutions, the most widely-used types of databases are SQL-oriented. However, this doesn't mean that a NoSQL can not be used. The main difference lies in the reason they were designed in the first place and a perfect, complete BI solution would be to use both for their own specific purpose.

However, it lacks the approaches to the NoSQL data warehouse. Due to the support of the parallel and distributed computing framework, MapReduce can be used to improve the performance of the creation of the NoSQL data warehouse.

I'd invite you to check out our own solution - the Databasel . It was designed as Business Intelligence for NoSQL (MongoDB is supported).