{"id":16197,"date":"2026-01-13T11:19:18","date_gmt":"2026-01-13T10:19:18","guid":{"rendered":"https:\/\/www.uni.lu\/research-en\/?post_type=events&p=16197"},"modified":"2026-02-16T11:50:41","modified_gmt":"2026-02-16T10:50:41","slug":"doctoral-defence-sangeeta-kakati","status":"publish","type":"events","link":"https:\/\/www.uni.lu\/research-en\/events\/doctoral-defence-sangeeta-kakati\/","title":{"rendered":"Doctoral Defence: Sangeeta KAKATI"},"content":{"rendered":"\n
<\/div>\n\n\n

The Doctoral School in Science and Engineering<\/a> is happy to invite you to Sangeeta KAKATI\u2019s defence entitled<\/p>\n\n\n\n

\nCross-Architecture Application Portability Using WebAssembly in Distributed Edge-Cloud Systems<\/h2>\n\n\n\n

Supervisor<\/strong>: Dr. Mats BRORSSON<\/a><\/p>\n\n\n\n

Cloud computing has transformed the IT landscape by enabling elastic resource provisioning with minimal upfront cost. However, as distributed and client-facing applications increasingly migrate toward the network edge, the once homogeneous cloud has evolved into a fragmented ecosystem spanning diverse architectures and accelerators. Developers now face the challenge of deploying and efficiently executing applications across heterogeneous platforms without detailed knowledge of underlying hardware. This growing complexity demands a portable and architecture-agnostic execution model that bridges the cloud\u2013edge continuum.<\/p>\n\n\n\n

In this thesis, we explore WebAssembly (Wasm) as a unifying abstraction for heterogeneous and distributed environments. We conduct a comprehensive review of WebAssembly beyond the web, identifying key challenges related to runtime efficiency and cross-architecture interoperability. Building on this foundation, we perform an extensive performance study of runtimes across X86_64, ARM64, and RISC-V platforms, providing a detailed cross-architecture characterization in terms of startup latency, execution speed, and compiler trade-offs.<\/p>\n\n\n\n

We further examine the role of WebAssembly in containerized infrastructures by comparing Wasm-based containers with traditional containers under Docker and containerd runtimes. Our results show that Wasm containers achieve up to 70% smaller image sizes and 25% faster cold pulls while maintaining comparable startup times, establishing WebAssembly as a compact and efficient alternative for multi-platform deployment.<\/p>\n\n\n\n

Finally, we investigate the extensibility of the WebAssembly System Interface (WASI) through experimental host-side services implemented in Wasmtime. We design and implement prototypes that introduce capability-based abstractions for host-managed memory and data-centric operations, including dataframe-style analytics. Through these prototypes, we demonstrate how components can interact with host-provided services using handle-based interfaces, illustrating a viable design path for offloading computation beyond the Wasm sandbox.<\/p>\n\n\n\n

Overall, we contribute to both theory and practice by (i) establishing empirical foundations for understanding WebAssembly performance across heterogeneous systems, (ii) validating its potential as a portable runtime for containerized applications, and (iii) providing an initial exploration of how WASI can be extended toward host-assisted and data-aware execution. The findings highlight WebAssembly\u2019s promise as a sustainable, cross-platform substrate for next-generation cloud\u2013edge computing.<\/p>\n\n\n\n