Nima Mahmoudi is a Machine Learning Engineer at Meta Platforms Inc. He received the BSc degrees in Electronics and Telecommunications and the MSc degree in Digital Electronics from Amirkabir University of Technology, Tehran, Iran in 2014, 2016, and 2017 respectively. He received the PhD degree in Software Engineering and Intelligent Systems from the University of Alberta, Edmonton, AB, Canada in 2022. He used to be a visiting research assistant in the Performant and Available Computing Systems (PACS) lab at York University, Toronto, ON, Canada. Before joining Meta, he was a Machine Learning engineer on the AI Accelerator team at TELUS.
PhD in Software Engineering and Intelligent Systems, 2022
University of Alberta
MSc in Digital Electronics, 2017
Amirkabir University of Technology
BSc in Telecommunications, 2016
Amirkabir University of Technology
BSc in Electronics, 2014
Amirkabir University of Technology
In our Serverless Computing research, our goal was to characterise and optimize the serverless computing platforms, making them adaptive and improving their performance.
In our last research in Performance and Available Computing Systems (PACS) Lab at York University, we were building accurate and tractable performance models for metric-based serverless computing platforms like Knative and Google Cloud Run that can estimate cost and performance of a given workload under different conditions accurately. Our paper titled “Performance Modeling of Metric-Based Autoscaling in Serverless Computing” is still being pending review, but our research has shown promising results, both in terms of accuracy and applicability.
You can check out our latest publications on our website.
In a previous research, we used Queuing Theory and Semi-Markov Processes to build a performance model for serverless computing platforms, with adequate fidelity and tractability to be used for modelling large-scale deployments. Our paper in this research called “Performance Modeling of Serverless Computing Platforms’’ was published in IEEE Transactions on Cloud Computing (TCC). In that paper, we were able to successfully model the performance metrics of AWS Lambda. We were also able to identify a key configuration in serverless computing platforms that can be used to optimize the overall performance of the system.
In a previous study we modelled the transient aspect of the key performance metrics of modern serverless computing platforms and showed its accuracy on AWS Lambda. This work was published in the Sixth Workshop on Serverless Computing (WoSC'20) as part of the ACM Middleware conference. This work was an extension to our steady-state performance modelling paper which was published in IEEE Transactions on Cloud Computing (TCC).
In another previous study in Dependable and Distributed Systems Lab (DDSL) at the University of Alberta, we used workload profiling and machine learning to build an adaptive function placement algorithm for serverless computing platforms. The resulting algorithm was able to achieve a better performance than the state of the art using similar hardware to perform the computing tasks.
My latest research in Computer Vision in Signal and Speech Processing Research Lab (SPRL) was on Multi-Target Object Tracking to produce high-quality object trajectories while maintaining low computational costs in order to be applicable for live implementation (e.g., autonomous vehicles). This research led to my master’s thesis and a journal paper titled “Multi-target tracking using CNN-based features: CNNMTT’’ published in Multimedia Tools and Applications. In this work we used the mid-layer output of custom-made Convolutional Neural Networks (CNN) as visual queues which combined with formal tracjectory models created smooth and stable tracking results. The resulting method was one of the top 4 methods in MOT Challenge Benchmark at the time.
In a previous research we used computer vision in Control of Multi-Vehicle Systems (CMVS) Lab in the field of robotics. Some of the projects completed between 2012-2017 included using computer vision for controlling Unmanned Ground Vehicles (UGVs) and Quadcopters and robust hand tracking using LBP features and Kalman Filters.
Serving machine learning inference workloads on the cloud is still a challenging task on the production level. Optimal configuration of the inference workload to meet SLA requirements while optimizing the infrastructure costs is highly complicated due to the complex interaction between batch configuration, resource configurations, and variable arrival process. Serverless computing has emerged in recent years to automate most infrastructure management tasks. In this work, we present MLProxy, an adaptive reverse proxy to support efficient machine learning serving workloads on serverless computing systems.
Analytical performance models are very effective in ensuring the quality of service and cost of service deployment remain desirable under different conditions and workloads. While various analytical performance models have been proposed for previous paradigms in cloud computing, serverless computing lacks such models that can provide developers with performance guarantees. In this work, we aim to develop analytical performance models for the latest trend in serverless computing platforms that use concurrency value and the rate of requests per second for autoscaling decisions.
Developing accurate and extendable performance models for serverless platforms, aka Function-as-a-Service (FaaS) platforms, is a very challenging task. Also, implementation and experimentation on real serverless platforms is both costly and time-consuming. However, at the moment, there is no comprehensive simulation tool or framework to be used instead of the real platform. As a result, in this paper, we fill this gap by proposing a simulation platform, called SimFaaS, which assists serverless application developers to develop optimized Function-as-a-Service applications in terms of cost and performance.
In this work, we propose an analytical performance model that captures the unique details of serverless platforms. The model can be leveraged to improve the quality of service and resource utilization and reduce the operational cost of serverless platforms. Also, the proposed performance model provides a framework that enables serverless platforms to become workload-aware.