Author Topic: Microsoft's Azure - Anyone using this???  (Read 279 times)

0 Members and 1 Guest are viewing this topic.

Offline NW27

  • Contributor
  • Posts: 48
  • Country: au
Microsoft's Azure - Anyone using this???
« on: September 10, 2019, 01:57:17 am »
Hi All,

I'm creating a IoT device that will need to connect to the cloud for data storage.
I'm looking at using Microsoft's Azure and an SQL database.

Anybody else do this that can ease my learning curve a bit :) ?

Thanks,
Neril.
 

Offline Fred27

  • Supporter
  • ****
  • Posts: 550
  • Country: gb
    • Fred's blog
Re: Microsoft's Azure - Anyone using this???
« Reply #1 on: September 10, 2019, 10:15:23 am »
Three documentation on Azure is very good. Take a look at Microsoft's examples.
 

Offline obiwanjacobi

  • Frequent Contributor
  • **
  • Posts: 915
  • Country: nl
  • What's this yippee-yayoh pin you talk about!?
    • Marctronix Blog
Re: Microsoft's Azure - Anyone using this???
« Reply #2 on: September 12, 2019, 12:50:02 pm »
Let your device call a function with the payload.
The function can then connect to the SQL db and transform and store the data.

If you use something like Table storage, it is way cheaper. An Azure SQL DB is not cheap.
Table storage can also be called directly.

[2c]
Arduino Template Library | Zalt Z80 Computer
Wrong code should not compile!
 

Offline NANDBlog

  • Super Contributor
  • ***
  • Posts: 4437
  • Country: nl
Re: Microsoft's Azure - Anyone using this???
« Reply #3 on: September 12, 2019, 01:09:13 pm »
You need some service running on the server, that acts like the endpoint for the communication. Like an MQTT broker. That places the data in the SQL database.  Stuff like this:
http://www.steves-internet-guide.com/logging-mqtt-sensor-data-to-sql-database-with-python/

The Azure server can be just a VM, doing the computing, or you can have just a database access, running on the cloud. One way, you are basically a sysadmin of the VM, and you can just develop the thing on a local linux machine, and then place it on the cloud. This can be easy to set up and develop, and then depending on your requirements, could be enough for years.
And then you can ask them to "give me a database", and they can take care of everything. Could cost more, but if you suddenly end up with 50 million users, you dont have to worry about database scaling. If you dont expect that, just buy a VM. Or you know, several VMs all doing different parts of the entire back/frontend. Whatever makes sense.
 

Offline obiwanjacobi

  • Frequent Contributor
  • **
  • Posts: 915
  • Country: nl
  • What's this yippee-yayoh pin you talk about!?
    • Marctronix Blog
Re: Microsoft's Azure - Anyone using this???
« Reply #4 on: September 12, 2019, 01:15:00 pm »
Azure is most effective if you use its PaaS services. So trying to do it yourself in a VM is 'suboptimal'.

You can get an idea what it'll cost you here: https://azure.microsoft.com/en-us/pricing/calculator/

Arduino Template Library | Zalt Z80 Computer
Wrong code should not compile!
 

Offline ehughes

  • Frequent Contributor
  • **
  • Posts: 359
  • Country: us
Re: Microsoft's Azure - Anyone using this???
« Reply #5 on: October 07, 2019, 06:01:56 pm »
Look into Azure IOT Hub.   It has front end interfaces for MQTT, AMPQ and HTTPS  for ingesting data with a security model ready to go.  I believe there is already event triggers to get data directly into a DB.    You can also have an Azure function triggered (which can be written in C#, javascript or python) to deal with the ingestion.

I have been using Redis quite a bit for caching data before going to a cold storage DB.

If you are recording TimeSeries data,   a SQL DB is often the worst way to store data.     Since almost all data will be inserts,  you need to be smart about partitioning if you plan to scale.    A transitional RDBMS stores data in a way that is not conducive to scaling time series data. (Related to the be b-tree).    If you are staying in Azure,  Looking in Cosmos or Azure tables.    I saw a 10x performance increase for an equivalently priced service.        The key is to 1st understand how you plan to partition, store and query your data. The DB technology will follow suite.

Technologies suchas InfluxDB do a good job for storing time series.   If this is a one-off then forget everything I just said and get it work anyway you can.  If this has a any scale factor (i.e. thousands of devices) then you need to think about this ahead of time.

 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf