AWS Builder Center

Argon: Drag-and-drop AWS architecture simulator

Drag-and-drop AWS architecture simulator. Build it, load test it, review it, deploy it. with AI autocomplete

 Argon  is a simulator with a graph editor that deterministically simulates AWS architectures under load. You place services such as EC2, Aurora, Lambda, SQS and CloudFront on a canvas, connect them, set the number of users, and read latency, throughput and dropped requests for each node.
We made Argon as part of the AWSxWeMakeDevs First Commit hackathon. If you checked out the project and liked it, leave us a star over on  Argon's GitHub repo .

The problem it solves

You usually find out an architecture is undersized after the traffic arrives, when the fix costs an incident. Argon lets you find the bottleneck first: raise the user count until a node saturates, then change the instance type or add capacity and run it again. For an AWS developer, that means answering "does this hold at 50k users, and what does it cost?" in minutes, with no deployment.
Argon models real AWS services, while most simulators are generic and do not allow you to config params


Specifically AWS

General system design simulators model abstract boxes like "database" and "cache." We wanted to see the consequences of decisions made at a finer level: which instance family, which storage tier, which region, which scaling limits. Argon models those.
EC2 nodes use instance-type and pricing-plan data, S3 uses storage tiers and cost, and the Region frame uses inter-region latency. Auto Scaling Groups take min and max limits, so scale-out and scale-in show up in the run.

How it works

A whole lot of data. A few mathematical models ( 1 ,  2  and more) and a graph storing the connections between nodes.
Each service has its own throughput, latency and drop model, and a request's path through the graph is computed hop by hop, so a saturated node affects everything downstream.
Frames for VPCs, regions and ASGs group nodes and apply their constraints (i.e region-wise latency, ASG scaling periods etc. are all applied).
We used Strands Agents to make LLM generate architecture graphs as structured output, so the AI autocomplete can read your graph, predict what and where you would like to place things, and connect them to each other. Strands' session management features ensure that all completions, suggestions and your input are part of the same conversation, so you don't have to provide context again.

Walkthrough



Inspiration

  •  Breakscale : build a system, load it until it breaks, watch why
All of the above are generic and cloud-agnostic, however Argon adds AWS-level detail such as instance types, pricing and regions. + AI autocomplete too.
Try it out at:  Argon 
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article