3 ms·

Probably going to get downvoted for this, but I feel like MongoDB should get more love in these subs. We use it all the time and their aggreations can get real

Probably going to get downvoted for this, but I feel like MongoDB should get more love in these subs. We use it all the time and their aggreations can get really advanced and perform well to the level where we run most analytics on demand. Sure we're not pushing to “big data” levels, max a few 100k records, but in reality I believe that's the average size of the majority of business datasets (Just an estimate, I have nothing to back this up) Been building with it for 5 years now and it's been a breeze. Especially with Atlas. I think our team has not spent more that 3 days in total on DB dev ops. And with Atlas Lucene text search and data lakes, querying data from S3. What's not to love.

jmchuster 5y ago
Yes, if you're only going to have 100k records, then basically any type of store will work, and you should choose purely on ergonomics. At that size, you can get away with not even having indexes on a traditional sql database.
ahmedelsama 5y ago
Yeah, for this blog, it was optimizing a Postgres table with 100m records. so 1000x more and thus all these issues came to be.
adwww 5y ago
Mongo is nice to work with for applications, but for analytics it's a bit of a nightmare - mostly around non compatibility with external tools.
neximo64 5y ago
I have used Mongo and it is a total nightmare, a bit of the reasons are in your text. Its supposedly only good with Atlas which is a managed service. I haven't tried Atlas myself (and why would I? - I try and avoid lock in) but since Postgres supports json column types which has been my go to instead of Mongo & it has been an absolute breeze. Especially since it can be indexed and scanned with postgres sql.
arcticfox 5y ago
> Sure we're not pushing to “big data” levels, max a few 100k records, but in reality I believe that's the average size of the majority of business datasets (Just an estimate, I have nothing to back this up) Excel on a laptop is also a viable option at that scale
KptMarchewa 5y ago
Even not large banks/fintech often add more records daily.
Jweb_Guru 5y ago
In analytic workloads MongoDB gets absolutely eviscerated on performance, it is absolutely not optimized for OLAP. Obviously, you've got to go with what you know and it's great that it works for you (though at your data sizes, I'm pretty sure you could also perform quite well by just loading all the data into a vector in memory in your application, or maybe even put it in Excel as another comment mentions :)), but generally speaking this article is aimed at people whose data is already in Postgres or are familiar with it and want to leverage it for OLAP. If that's the case, putting that data in MongoDB and performing worse on basically all fronts isn't a very compelling idea.
beckingz 5y ago
You are correct that most business datasets are small. However, once you get to a 600 M row dataset, you will likely encounter significant performance/cost issues with a noSQL database.

jmchuster 5y ago

Yes, if you're only going to have 100k records, then basically any type of store will work, and you should choose purely on ergonomics. At that size, you can get away with not even having indexes on a traditional sql database.

ahmedelsama 5y ago

Yeah, for this blog, it was optimizing a Postgres table with 100m records. so 1000x more and thus all these issues came to be.

adwww 5y ago

Mongo is nice to work with for applications, but for analytics it's a bit of a nightmare - mostly around non compatibility with external tools.

neximo64 5y ago

I have used Mongo and it is a total nightmare, a bit of the reasons are in your text. Its supposedly only good with Atlas which is a managed service. I haven't tried Atlas myself (and why would I? - I try and avoid lock in) but since Postgres supports json column types which has been my go to instead of Mongo & it has been an absolute breeze. Especially since it can be indexed and scanned with postgres sql.

arcticfox 5y ago

> Sure we're not pushing to “big data” levels, max a few 100k records, but in reality I believe that's the average size of the majority of business datasets (Just an estimate, I have nothing to back this up) Excel on a laptop is also a viable option at that scale

KptMarchewa 5y ago

Even not large banks/fintech often add more records daily.

Jweb_Guru 5y ago

In analytic workloads MongoDB gets absolutely eviscerated on performance, it is absolutely not optimized for OLAP. Obviously, you've got to go with what you know and it's great that it works for you (though at your data sizes, I'm pretty sure you could also perform quite well by just loading all the data into a vector in memory in your application, or maybe even put it in Excel as another comment mentions :)), but generally speaking this article is aimed at people whose data is already in Postgres or are familiar with it and want to leverage it for OLAP. If that's the case, putting that data in MongoDB and performing worse on basically all fronts isn't a very compelling idea.

beckingz 5y ago

You are correct that most business datasets are small. However, once you get to a 600 M row dataset, you will likely encounter significant performance/cost issues with a noSQL database.