Muna On The Road: Optimizing Network Edge Processing via oneAPI

Happy New Year, everyone! I hope your 2026 is off to a fantastic start.

It has been a couple of months since I last posted, but a lot has happened. To share some quick updates, I have officially started my PhD program in Computer Science at Cornell University, where my research focuses on Computer Systems and Networks. My goal for this next chapter is to take my understanding of the field to the deepest possible level, exploring the details that go into designing, building, and optimizing the infrastructure that powers our world.

As I embark on this academic journey, I view my speaking engagements and this blog as vital avenues to share my experience with you. I want to bridge the gap between advanced research and practical application, bringing you along for the ride as we explore new frontiers in technology.

With that, here is the latest stop on my "Muna On the Road" series!

I had the privilege of speaking at the Unified Acceleration (UXL) Foundation's Developer Summit: oneAPI DevSummit Hosted by UXL Foundation.

The Unified Acceleration (UXL) Foundation represents a major shift in how we think about hardware acceleration. While oneAPI began as an Intel initiative to create a unified programming model, it has evolved into something much larger. The project was moved under the UXL Foundation (hosted by the Linux Foundation) to ensure true vendor neutrality and open governance.

My session addressed a major hurdle in network operations: the "Data Overload" generated by modern telemetry, which often leads to delayed insights and missed security threats. Edge devices frequently suffer from limited computational power, making real-time analysis a significant challenge. My goal was to demonstrate how to bypass these limitations using the oneAPI Data Analytics Library (oneDAL).

During the talk, we covered three essential areas:

• The Edge Analysis Challenge: We explored why traditional CPU-based analysis struggles with the massive volume of modern telemetry data.
• Practical Anomaly Detection: We demonstrated how to use oneDAL to calculate covariance matrices on packet capture (.pcapng) data, allowing for the proactive identification of network irregularities.
• Measurable Performance Gains: Our experiments revealed that the oneAPI-accelerated code was approximately 2.71x faster than standard Python, proving the efficiency of offloading tasks to optimized hardware.

It was an exciting opportunity to show how vendor-neutral code can enable "on-box automation," allowing operators to run proactive analytics directly on network devices regardless of the hardware vendor.

Check out the full presentation below!