I agree with the advice you have been given to stick with SQL Server. If you are on the latest SQL Server version you can query inside the JSON field. You should set up a test database with a JSON field and try some queries. Once you understand it and can demonstrate it, show it to the other developers that are suggesting MongoDB. Once they see it working with their own eyes they may drop their position of Mongo over SQL. I would only seriously consider MongoDB if there was no other SQL requirements. I wouldn't do both. I'd be all SQL or all Mongo.

I think the key thing to look for is what kind of queries you're expecting to do on that JSON and how stable that data is going to be. (And if you actually need to store the data as JSON; it's generally pretty inexpensive to generate a JSON object)

MongoDB gets rid of the relational aspect of data in favor of data being very fluid in structure.

So if your JSON is going to vary a lot/is unpredictable/will change over time and you need to run queries efficiently like 'records where the field x exists and its value is higher than 3', that's a great use case for MongoDB.

It's hard to solve this in a standard relational model: Indexing on a single column that has wildly different values is pretty much impossible to do efficiently; and pulling out the data in its own columns is hard because it's hard to predict how many columns you'd have or what their datatypes would be. If this sounds like your predicament, 100% go for MongoDB.

If this is always going to be more or less the same JSON and the fields are going to be predictably the same, then the fact that it's JSON doesn't particularly matter much. Your indexes are going to approach it similar to a long string.

If the queried fields are very predictable, you should probably consider storing the fields as separate columns to have better querying capabilities. Ie if you have {"x":1, "y":2}, {"x":5, "y":6}, {"x":9, "y":0} - just make a table with an x and y column and generate the JSON. The CPU hit is worth it compared to the querying capabilities.

I am biased (work for Scylla) but it sounds like a KV/wide column would be better in this use case. Document/schema free/lite DBs data stores are easier to get up and running on but are not as scalable (generally) as NoSQL flavors that require a more rigid data model like ScyllaDB. If your data volumes are going to be 10s of TB and transactions per sec 10s of 1000s (or more), look at Scylla. We have something called lightweight transactions (LWT) that can get you consistency.

I have found MongoDB highly consistent and highly available. It suits your needs. We usually trade off partion tolerance fot this. Having said that, I am little biased in recommendation as I haven't had much experience with couchbase on production.

I prefer MongoDB due to own experience with migration of old archive of pdf and meta-data to a new “archive”. The biggest advantage is speed of filters output - a new archive is way faster and reliable then the old one - but also the the easy programming of MongoDB with many code snippets and examples available. I have no personal experience so far with Couchbase. From the architecture point of view both options are OK - go for the one you like.

I would like to suggest MongoDB or ArangoDB (can't choose both, so ArangoDB). MongoDB is more mature, but ArangoDB is more interesting if you will need to bring graph database ideas to solution. For example if some data or some documents are interlinked, then probably ArangoDB is a best solution.

To process tables we used Abbyy software stack. It's great on table extraction.

If you can select text with mouse drag in PDF. Use pdftotext it is fast! You can install it on server with command "apt-get install poppler-utils". Use it like "pdftotext -layout /path-to-your-file". In same folder it will make text file with line by line content. There is few classes on git stacks that you can use, also.

Hi Erin,

Honestly both databases will do the job just fine. I personally prefer Postgres.

Much more important is how you store the audio. While you could technically use a blob type column, it's really not ideal to be storing audio files which are "several hours long" in a database row. Instead consider storing the audio files in an object store (hosted options include backblaze b2 or aws s3) and persisting the key (which references that object) in your database column.

Hi Erin, Chances are you would want to store the files in a blob type. Both MySQL and Postgres support this. Can you explain a little more about your need to store the files in the database? I may be more effective to store the files on a file system or something like S3. To answer your qustion based on what you are descibing I would slighly lean towards PostgreSQL since it tends to be a little better on the data warehousing side.

Hey Erin! I would recommend checking out Directus before you start work on building your own app for them. I just stumbled upon it, and so far extremely happy with the functionalities. If your client is just looking for a simple web app for their own data, then Directus may be a great option. It offers "database mirroring", so that you can connect it to any database and set up functionality around it!

Hi Erin! First of all, you'd probably want to go with a managed service. Don't spin up your own MySQL installation on your own Linux box. If you are on AWS, thet have different offerings for database services. Standard RDS vs. Aurora. Aurora would be my preferred choice given the benefits it offers, storage optimizations it comes with... etc. Such managed services easily allow you to apply new security patches and upgrades, set up backups, replication... etc. Doing this on your own would either be risky, inefficient, or you might just give up. As far as which database to chose, you'll have the choice between Postgresql, MySQL, Maria DB, SQL Server... etc. I personally would recommend MySQL (latest version available), as the official tooling for it (MySQL Workbench) is great, stable, and moreover free. Other database services exist, I'd recommend you also explore Dynamo DB.

Regardless, you'd certainly only keep high-level records, meta data in Database, and the actual files, most-likely in S3, so that you can keep all options open in terms of what you'll do with them.

Hi Erin,

• Coming from "Big" DB engines, such as Oracle or MSSQL, go for PostgreSQL. You'll get all the features you need with PostgreSQL.
• Your case seems to point to a "NoSQL" or Document Database use case. Since you get covered on this with PostgreSQL which achieves excellent performances on JSON based objects, this is a second reason to choose PostgreSQL. MongoDB might be an excellent option as well if you need "sharding" and excellent map-reduce mechanisms for very massive data sets. You really should investigate the NoSQL option for your use case.
• Starting with AWS Aurora is an excellent advise. since "vendor lock-in" is limited, but I did not check for JSON based object / NoSQL features.
• If you stick to Linux server, the PostgreSQL or MySQL provided with your distribution are straightforward to install (i.e. apt install postgresql). For PostgreSQL, make sure you're comfortable with the pg_hba.conf, especially for IP restrictions & accesses.

Regards,

I recommend Postgres as well. Superior performance overall and a more robust architecture.