Architecture Diagram

[quartz zealot]

Hi, this is for a class project. I’m working on a model for PTEC Limited, a company that collects smart home sensor data and currently processes it on-premises. With growth, they expect to handle 500,000 data points per day, and their current servers won’t scale. The company wants to move to cloud-managed IoT services for processing, storage, and analytics.

Could you look at my model and let me know if this setup would help them meet their scalability and reliability goals?

[native mantle]

Have you tried sending that to an ai and asking it to find what is wrong with it

[spring lantern]

I tossed it at Claude on Opus Extended:

Okay, a few things jump out:

The good: The general flow makes sense — on-prem IoT devices → IoT Hub → stream processing → storage. Including Key Vault, MFA, and ExpressRoute/VPN shows they're thinking about security. The legend and boundary labels are nice for a class project.

The problems:

1. "Azure Stream Aelys" — it's Analytics. Typo, but in a class submission that's sloppy.

2. "Consainer" — appears twice in the labels. It's Container... or did they mean something else entirely? Unclear.

3. Two "Azure Data Lake Storage" boxes on the right side. Is that intentional (hot/cold tiers)? If so, label them differently. If not, it's a mistake.

4. No compute layer. Data flows from IoT Hub → Stream Analytics → storage... and then what? Where's the actual analytics/processing? There's no Azure Functions, no Databricks, no App Service, nothing that does anything with the data. For 500K data points/day that's not even a heavy load, but you still need something to act on it.

5. Azure Table Storage as a central node is a weird choice. For IoT time-series data, Cosmos DB (already in the diagram) or ADX (Azure Data Explorer) would be the primary store. Table Storage is legacy and limited in query capability. Why is it the hub?

6. The on-prem section is cluttered and unclear. There are icons for cameras, IoT devices, a firewall, a shield — but no labels explaining what they represent or how they connect. The big black arrows between servers mean... what?

7. Azure Migrate in the security section is odd. Migrate is a one-time migration tool, not an ongoing security/management component. It doesn't belong in that boundary long-term.

8. No monitoring. No Azure Monitor, no Log Analytics, no alerts. For a scalability/reliability goal, observability is non-negotiable.

Bottom line: The diagram shows awareness of Azure services but reads like someone browsed the Azure icon library and assembled components without fully understanding the data flow. The core pipeline (ingest → process → store → analyze → act) is incomplete — it stops at "store." For a class project it's a decent start, but it wouldn't pass a real architecture review. They need to clarify the compute/analytics layer and clean up the labeling errors.

Claude brings up valid points about the lack of compute and monitoring

[quartz zealot]

I haven’t added anything about monitoring yet, but I am adding Azure Monitoring, Azure Backup and Recovery, RBAC/Entra ID, Network Security Groups/Firewalls, and Azure Security. And I want you to look at this one to see if everythign is good.

[quartz zealot]

yes i have and thats what i used for the new one

[quartz zealot]

Thank you very much the things you mentioned helped

[spring lantern]

I can't speak for all services you have here but there are problems here.
Firstly, there are no labels.
As for this corner I cropped out, what is going on between Event Hubs, IOT Hubs, and Stream Analytics and why are the arrows all bidirectional?

[quartz zealot]

IoT Hub supports true two‑way device communication, Event Hubs handles high‑throughput one‑way ingestion, and Stream Analytics processes that incoming data in real time and outputs it to various destinations, so the bidirectional arrows simply represent multiple possible input/output paths rather than data flowing backward.

Labeled

[spring lantern]

Typically architectures would not have a single component both send and receive from the same Event Hub as that would be redundant. Usually multiple Event Hubs are involved, each one dedicated to a specific need.
I would recommend cleaning up that part so it makes more sense how that is used.

[quartz zealot]

I think I'm pretty much done all i need to do is the RBAC am pretty sure can you check for me?

the bottom part is security

[silver roost]

These screens are AI generated. Look at left top consainer

[native mantle]

I thought they just didn't understand anything and tried to build it together

[silver roost]

Why use Azure table storage

[native mantle]

Well you could make the argument that it is cheaper than cosmos db but they have that as well

[silver roost]

I'd rather use SQL Server instead of Azure tables

[spring lantern]

I chalked those up to maybe a language issue

But also uhmmmm the diagrams don't make sense

[silver roost]

No

Consainer

AI (still) can't write

[spring lantern]

I've tried some that are surprisingly good at it.

[silver roost]

This one wasn't 😂

@quartz zealot So please, write something yourself .

From now I won't allow any AI generated stuff in this thread on this topic

We expect some effort into questions.

[quartz zealot]

Bro I am useing drawio

[quartz zealot]

Also I am a student learning btw

[quartz zealot]

That is STILL learning why do you think I am asking for help.

[silver roost]

Hey, i only disallow AI.

[quartz zealot]

Where is the ai?

[silver roost]

And to be fair

We are already being generous

rule 12

This image is AI

Either way, we are happy ot help 🙂

[quartz zealot]

Yes that one is

[silver roost]

Tell me, why you want Auzre table service?

[quartz zealot]

I am not disagreeing

So I can store large structure data

That don't have anything to do together

Am pretty sure

[quartz zealot]

I pretty much don't and just started

[silver roost]

So what's the goal?

[quartz zealot]

The goal is to migrate PTEC Limited’s on-premises system to a scalable cloud-based IoT architecture that can handle increasing data volumes, improve performance, and support future business growth. They are going from 50k data point s a day to 500k

Am pretty sure I put the scenario in the chat I'm not sure tho

[silver roost]

So how did you come up with the above

Other then AI

[quartz zealot]

Do you mean why I chose those cloud services?

[silver roost]

Yes.

[quartz zealot]

Here the only thing I removed was azure event hub

[spring lantern]

The AI is telling you a lot of nothing.

https://learn.microsoft.com/en-us/azure/architecture/browse/?terms=iot

I think you should browse the official documentation on architecture patterns to find one that best fits your use case and then modify it to better fit.

https://learn.microsoft.com/en-us/azure/architecture/solution-ideas/articles/iot-azure-data-explorer

Might be a fit for you. But only you can tell. We can't determine it for you.

[quartz zealot]

Alright thank you

[silver roost]

I agree with Ziyan 🙂

[spring lantern]

Additionally, if you just must use AI, I would recommend having it start by using the documented architectures as a base and then apply modifications.
I would still recommend learning the architectures yourself so you can better understand the role of each component and the why behind the architectural decisions of each component.

Keep in mind that in a real world setting each component comes with a cost and someone has to support/maintain it. As a school exercise it is trivial to add/remove unnecessary components but real world application is nontrivial when downtime costs the company lots of money.

[quartz zealot]

Ya I also have the cost documented by the word doc

[haughty forum]

Would there be an opportunity to work together as a software engineer?

[silver roost]

With the goal of asking money ;)?

[haughty forum]

Do you have any project idea?

[native mantle]

It's a bot

[spring lantern]

He's just playing with his prey lol