Accelerated Computing Systems Group

We are a part of the Systems Research Group, focusing on designing and building a secure, efficient, and high-scalable end-to-end computer system by adopting emerging hardware devices. Our research projects leverage a wide range of state-of-the-art hardware technologies/devices, such as accelerators (FPGAs, GPUs, ASICs), SmartNICs, Infrastructure Processing Units (IPUs), trusted execution environments (Intel SGX/TDX, ARM TrustZone, AMD Sev), silicon root-of-trust (OpenTitan), targeting modern cloud computing infrastructures, such as cloud-native environments, serverless computing, distributed systems, resource-disaggregated servers, embedded/IoT systems.

NOTE: we’re looking for students who want to write BSc/MSc theses or participate in Guided Research. If you are interested, please send us your application by following our application instructions.


Ongoing Research Projects

Trustworthy Disaggregated Architectures

The rising performance demands and increasing heterogeneity in cloud data centers lead to a paradigm shift in the cloud infrastructure, from monolithic servers to a disaggregated architecture. In a multi-tenant cloud, users should be able to leverage trusted computing to protect their applications from untrusted parties. While Trusted Execution Environments (TEEs) are a well-known technique for realizing trusted computing on monolithic servers, we cannot adopt existing TEE technologies to the disaggregated architecture due to their distributed nature and heterogeneity of devices.

This project aims to design and build trusted heterogeneous disaggregated architectures that allow cloud users to construct virtual TEEs (vTEEs): TEE-based, secure, isolated environments assembled with any combination of disaggregated components.

Keywords: Resource disaggregation, Heterogeneous computing, Trusted Execution Environments (TEEs), Intel IPUs, OpenTitan

Publications: APSys’23 (paper, slides)

Cloud-native FPGA Virtualization

Field Programmable Gate Arrays (FPGAs) are being adopted in cloud-native environments as they are suitable for accelerating modern cloud workloads: machine learning, databases, graph processing, and so on. Despite their advantages, there is currently no distributed FPGA virtualization support in cloud-native environments, which complicates on-demand FPGA management across distributed nodes, such as preemptive scheduling and migration, leading to low resource utilization and high running costs.

This project aims to propose and build a new cloud-native architecture for virtualizing and orchestrating distributed FPGAs in the cloud. Our system leverages a lightweight library OS (unikernel) to facilitate virtualizing distributed FPGAs with negligible performance overheads and realizing pre-emptive task scheduling for FPGA resource orchestration.

Keywords: Cloud-native environments, Microservices, FPGAs, Virtualization, Unikernels

FPGA-accelerated Serverless Computing

Serverless computing has gained popularity in recent years due to its ease of use, scalability, and cost-effectiveness. At the same time, the increasing complexity of modern workloads has created a pressing need for hardware acceleration in the cloud and serverless computing. FPGAs are suitable for hardware acceleration in serverless environments, offering high flexibility, low latency, and high throughput.

This project aims to build an end-to-end serverless framework that facilitates the use of FPGAs in serverless computing architectures. Users can simply deploy or use FPGA functions without worrying about FPGA device management. Our system handles the execution of FPGA functions, focusing on resource consolidation, isolation, high throughput, and low latency.

Keywords: Serverless computing, FPGAs, Hardware acceleration, Kubernetes

I/O Acceleration Framework for FPGA-based Smart Devices

Modern cloud systems have adopted a variety of FPGA-accelerated I/O devices, such as SmartNICs and computational storage, while they face programmability and portability challenges. Existing FPGA frameworks either directly expose device-specific I/O interfaces to user logic or offer virtualized I/Os limited to a single device type. The lack of I/O abstraction imposes high engineering costs, less design portability, and even unexpected throughput degradation.

This project aims to build an FPGA-based I/O acceleration framework that brings better programmability and design portability. The framework offers a new FPGA Shell, a set of dedicated hardware modules to abstract FPGA fabrics, that virtualizes input/output ports of user logic and abstract device-dependent I/O specifications from its hardware design. Host applications can easily configure the connectivity between virtualized I/O ports and external I/O devices through simple APIs.

Keywords: FPGAs, SmartNICs, SmartSSDs, Near-data processing, I/O offloading

Publications: USENIX ATC’24 (to appear)


Group Members

Dr. Atsushi Koshiba

Research Group Leader

Jiyang Chen

PhD student

Harsha Unnibhavi

PhD student

Teofil Bodea

PhD student

Felix Gust

Research Engineer

Julian Pritzi

Student Assistant


Ongoing & Past MSc/BSc theses, GRs, IDPs

You can find reports, theses, and final presentations of finished projects in our research work archive.

StudentTypeTitleAdvisersEnd date
Martin LambackMScUtilizing dynamic partial reconfiguration in an FPGA-accelerated FaaS architectureCharalampos Mainas, Atsushi Koshiba2024, Summer (ongoing)
Felix GustMScPrototyping a Secure Controller for Trusted Heterogeneous Disaggregated ArchitecturesAtsushi Koshiba2023, Summer
Bruno ScheuflerBScCloud-Native Scheduling for Serverless FPGAsCharalampos Mainas, Atsushi Koshiba2023, Summer
Anand Krishna RallabhandiMScHeterogeneity-Aware Scheduling Algorithms for FPGA Workloads in Cloud EnvironmentsAtsushi Koshiba2023, Summer
Jiong LiuBScEvaluation of Data-Intensive Accelerated Functions in Computational Storage DevicesCharalampos Mainas, Atsushi Koshiba2022, Winter
Yuan LuoGRA Comprehensive Study of Coyote Memory SystemAtsushi Koshiba, Jiyang Chen2022, Winter
Julian PritziGRSecuring Hardware Communication using Encryption and AttestationHarshavardhan Unnibhavi2022, Winter
Zirong CaiMScFrontend Development of Distributed FPGA Management in Serverless ComputingAtsushi Koshiba, Jiyang Chen2022, Winter
Julian PritziBScAn in-hardware cycle-accurate benchmarking tool for security-critical operationsHarshavardhan Unnibhavi2022, Summer