Lots of improvements on this v0.17 release: some are “invisible”, but mission critical, such as the platform improvements on security and scalability, others you will notice right away, such as the new awesome Benchmarks Dashboard or the Tasks’ System Metrics.

Below, we’ll dive deeper into how to use these features to help you run simulations more efficiently and cost-effectively, and also breakdown when and why to use each of them.

Benchmarks Dashboard

We’ve said it before, but we can’t repeat it enough: Benchmarking is a “super power” tool that can save you a lot of time AND money!

It was already possible to run benchmarks from our Python client, which allows users to easily run a simulation on a set of cloud machines, and find out the running times and cost for each of them. Typically this is done in a simulation that is fully representative of what you care about, but with a much smaller number of steps (say 1%) — or when you know that later you will need to run hundreds or thousands of variations of the same simulation, with different parameters. This allows users to pick what they need: either the fastest machine (if they are in a hurry!) or the cheapest (if they are price sensitive and will run many simulations), or even a machine with a good trade-off between cost and speed.

The real-time part of the dashboard shows the progress while the benchmark is running, including overall progress, the list of tasks, and the machine groups involved in the benchmark.

Once it is finished, you will be able to see some beautiful plots showing execution duration per machine type, estimated cost per machine type and overall recommendation score. You’ll also see insightful data on the fastest and cheapest machine types, plus Inductiva’s recommended machine – that balances both speed and cost!

System Metrics – Fit the Machine To Your Simulation

We’ve added a new section to the Task Detail screen: System Metrics. This shows you key performance stats for the machine that ran your task — including CPU and memory usage, plus total disk read and write.

Please note that these metrics are not live; they become available only after your task finishes.

With System Metrics, you can better understand how your simulation used the computational resources, helping you fine-tune machine choices for faster runs and lower costs.

With the system metrics feature, you can gain valuable insights into your task’s performance and efficiency, for example:

Debug Performance Issues: Analyze the graphs to understand why a task was slow.
- A task constantly hitting 100% CPU usage is likely CPU-bound, so the user should consider running it on a machine with a higher number of vCPUs.
- A failed task whose RAM usage was very high might have failed due to lack of memory, so it could be worthwhile to try running it on a machine with a higher memory profile.
- A steady increase in memory usage might indicate a memory problem.
Identify Resource Bottlenecks: Easily determine if your task’s performance is being limited by CPU, memory, or disk I/O, allowing you to address the specific constraint.
Optimize Resource Allocation: Make data-driven decisions for future runs. If a task only uses a small fraction of the allocated memory and CPU, you can select a smaller, more cost-effective machine next time.

For extra guidance on picking the right cloud machine for your simulation, look into our guide

If you’re asking yourself when and why to use Benchmarks and System Metrics, here’s our rationale – link.

Security

We allow a lot of flexibility on our platform, even bringing your own software, using custom containers, so we need to make sure that if a user tries to abuse the platform, it will not be possible to impact other users (and we will quickly block their account!). To achieve that we made sure that any task that is submitted runs on a containerized environment without access to external resources (e.g. internet). We also protected our API admin endpoints via a combination of two very different authentication mechanisms, which means that even if one of them would be compromised, it would still not be possible to impact other user accounts.To validate the robustness of our security measures, we recently underwent a Red Teaming exercise conducted by Critical Software, a reputable and worldwide cybersecurity firm. During this assessment, security professionals attempted to compromise our API on a controlled replica of our production environment, ensuring no real user data was exposed.

We were glad that despite the attempts our API stood still without being compromised!

Obviously, security is not something you “achieve” and then can stand still: it is an ongoing process of improvement. It requires constant monitoring, and the adoption of the latest best practices and technologies. This is what we are committed to do, to meet the needs of our customers.

Scalability

A programmable platform for engineers and scientists to run their numerical simulations on the cloud needs to be scalable. Our current quotas allow business and academia plans to launch up to 100 Virtual Machines simultaneously, and each of them can have dozens of CPU cores. To make sure we can have many users running simulations in parallel, we recently made a series of performance improvements on our backend. Basically with the same amount of Google Cloud run containers of our web API we can now sustain nearly 20x more VMs running simulations at the same time, which means that many more users can enjoy the platform without being affected by slow response times.

Email Alerts of Low Credits

When users are close to running out of credits, we now send an alert by email, reminding them to top-up their account so that their running simulations are not interrupted.

NOTE: when a task is interrupted due to lack of credits, the outputs generated until that point are still saved in the user’s cloud storage.

For more detail on how the credits are used, have a look at our guide

Task Details, Streamlined & Simple!

The new tabbed layout in the Task Detail Console screen keeps everything neatly organized, so you spend less time scrolling and more time finding the data you need, faster.

New Project Overview — More Insights, Less Effort

The redesigned Project Detail screen gives you a clearer view of your project’s health and resources. You’ll now see helpful metrics like success rate and average task duration at a glance, plus easy access to the project’s Machine Groups — all alongside the Tasks table.

New Simulators

Last, but not least, we added new Simulators to our growing collection:

WRF (Weather Research and Forecasting Model): The Weather Research and Forecasting (WRF) Model is a state-of-the-art numerical weather prediction system developed to support both atmospheric research and operational forecasting. It provides a flexible, scalable platform capable of simulating a wide range of meteorological phenomena — from global-scale patterns to localized events like thunderstorms and hurricanes.

Dive into our comprehensive guide covering everything you need to know about WRF at Inductiva

SFINCS (Super-Fast INundation of CoastS): The SFINCS model is a reduced-complexity simulation engine developed by Deltares to simulate compound coastal flooding with high computational efficiency while maintaining reliable accuracy. With its grid-based structure that handles complex topographies, SFINCS is especially well-suited for large-scale, high-resolution inundation modeling where speed is critical.

Delve into the detailed guide we created for all things SFINCS at Inductiva

CM1 (Cloud Model 1): is a three-dimensional, time-dependent, non-hydrostatic numerical model designed for idealized studies of mesoscale atmospheric phenomena. Researchers use it in particular to study deep convective precipitation (thunderstorms).

Have a look into the guide we put together for CM1 at Inductiva

Our Insights