Hi,
I have read that hybrid relational and nonrelational databases are a good approach for IoT applications where real time analytics are needed (correct me if I'm wrong)
For example, MySQL (relational) and MongoDB (nonrelational)
Which is your opinion about this and there is some paper or information somewhere where I can read more about this?
New to databases here, all information is welcome.
Regards,
Here is my short answer to your short question: Time Series database, if not that, then NoSQL type database. Here is the long answer.
It depends. I am sure my peers may have different opinions on this, so I qualify my response as the simple opinions of a DBA who has started to enter the IOT arena from a business point of view, vs. scientific. I present to you a long rant, but you did ask about databases and I am a dba.
Closer examination of your requirements will bring you closer to a solution to consider. For example, as a peer previously stated, there are choices of database TYPES - relational, NoSQL and I add, Time Series. Each type has it's own advantages but no one is better than the other, hence your own requirements will lead you.
Generally for IOT, you want something lightweight (not much overhead and does not require significant resources), that can accommodate many small transmissions, over and over again, so as to become large by volume. I would start by considering the time series database: https://en.wikipedia.org/wiki/Time_series_database, which is more process efficient than the traditional relational database. You really do not need relational databases for IOT unless you are going to take advantage of relational specific functionality, as relational databases require you to send data in a stringent (perfect) format to match the record definition in the database.
NoSQL databases, do not require you to stick to a predefined or stringent format to insert data. Just send what you have and in it goes. To cross reference the difference between the relational and NoSQL database goes something like this: I send first_name, last_name, address and zipcode into a NoSQL database. In it goes...The next record I send to the NoSQL database only has first_name and last_name. In it goes to. Try the same thing with a relational database and the second record will be rejected as the data is incomplete. The trade off is, in the NoSQL database, the second record will never show up on a query (search) using address.
So I close with a few requirement questions, which I hope can bring you closer to what you are seeking:
1. What kind of data are you collecting? Is it vital to your business?
2. How important is each piece of data you collect? It's it monitoring life-support or simply the temperature of a pair of sneakers on a pallet?
3. Which component does the actual data interpretation? Asking if the database is to make sense of the data or simply serve as a collection repository for an application program to extract the data for processing.
4. What happens if you lose the database? Can you easily recover or do you have to start over. Terms such as High-availability and Disaster recovery are common considerations for database processing. A single silo of data is a single point of failure. Whether or not your choice of databases offers redundancy should at least be considered.
4. How often do you send the data? Who knows, if the data is collected and bundled before sending to a database in a burst, then maybe a relational database has value, then again...
--Best of luck,
John
i think NoSQL is better for big data storage, so i would recommend it for big sensor networks
https://www.tutorialspoint.com/mongodb/index.htm
Here is an absolute beginner resource to MongoDB, worth reading if you have no prior knowledge of it.
Hi,
For IoT, you may also have a look at InfluxDB and related stack:
https://www.influxdata.com/time-series-platform/influxdb/
Blockchain also plays an important role in a range of IoT application scenarios. You may wish to begin experimenting with BigchainDB (or similar) to learn more about the core concepts: https://www.bigchaindb.com/
Here is my short answer to your short question: Time Series database, if not that, then NoSQL type database. Here is the long answer.
It depends. I am sure my peers may have different opinions on this, so I qualify my response as the simple opinions of a DBA who has started to enter the IOT arena from a business point of view, vs. scientific. I present to you a long rant, but you did ask about databases and I am a dba.
Closer examination of your requirements will bring you closer to a solution to consider. For example, as a peer previously stated, there are choices of database TYPES - relational, NoSQL and I add, Time Series. Each type has it's own advantages but no one is better than the other, hence your own requirements will lead you.
Generally for IOT, you want something lightweight (not much overhead and does not require significant resources), that can accommodate many small transmissions, over and over again, so as to become large by volume. I would start by considering the time series database: https://en.wikipedia.org/wiki/Time_series_database, which is more process efficient than the traditional relational database. You really do not need relational databases for IOT unless you are going to take advantage of relational specific functionality, as relational databases require you to send data in a stringent (perfect) format to match the record definition in the database.
NoSQL databases, do not require you to stick to a predefined or stringent format to insert data. Just send what you have and in it goes. To cross reference the difference between the relational and NoSQL database goes something like this: I send first_name, last_name, address and zipcode into a NoSQL database. In it goes...The next record I send to the NoSQL database only has first_name and last_name. In it goes to. Try the same thing with a relational database and the second record will be rejected as the data is incomplete. The trade off is, in the NoSQL database, the second record will never show up on a query (search) using address.
So I close with a few requirement questions, which I hope can bring you closer to what you are seeking:
1. What kind of data are you collecting? Is it vital to your business?
2. How important is each piece of data you collect? It's it monitoring life-support or simply the temperature of a pair of sneakers on a pallet?
3. Which component does the actual data interpretation? Asking if the database is to make sense of the data or simply serve as a collection repository for an application program to extract the data for processing.
4. What happens if you lose the database? Can you easily recover or do you have to start over. Terms such as High-availability and Disaster recovery are common considerations for database processing. A single silo of data is a single point of failure. Whether or not your choice of databases offers redundancy should at least be considered.
4. How often do you send the data? Who knows, if the data is collected and bundled before sending to a database in a burst, then maybe a relational database has value, then again...
--Best of luck,
John
I'm in the security business, so you can assume that my opinions are heavily biased. Both MySQL and MongoDB are notoriously insecure, and get hacked daily. I would vigorously oppose their use in any system I was designing.
Just for fun, why not design your own in-memory database?
https://www.researchgate.net/project/IDaaS-with-secure-data-at-rest/update/5b0360f74cde260d15e01961
Thanks for the answers, I was reading this:
https://www.wirelessdesignmag.com/blog/2015/03/database-management-iot-relational-nosql%E2%80%A6or-both
Regards,
- Badges
- Science topic