Server-side Wasm Applications: Replacing Docker with WASI

The presentation discusses the use of WebAssembly and edge computing for machine learning models to improve user experience and reduce network costs.

• WebAssembly and edge computing can be used to run machine learning models in the browser and reduce network costs
• Edge computing is ideal for running small models that are single-threaded and require low latency
• WebAssembly can improve user experience by reducing load times and allowing for more memory in the browser
• The content authenticity initiative uses machine learning models to fingerprint images

The speaker demonstrated the use of a machine learning model to fingerprint an image of Grace Hopper, which could be used to detect if the image had been altered in any way.

Adobe makes use of Wasm in its flagship web browser-based products including Photoshop, Lightroom, and Acrobat. This past year it has explored potential use cases for Wasm in the datacenter with wasmCloud. Of particular interest were the potential performance, cost, security, and compliance benefits. Wasm and WASI have many potential advantages over Docker and standard web frameworks in these areas, but what needs to be done to realize those benefits at Adobe? This presentation begins with a summary of Adobe's current use cases for Kubernetes, including areas in which server-side Wasm could offer significant benefits. It then proceeds to an exploration of Wasm/WASI platforms, compelling features of the technology for Adobe, and a demonstration of proofs of concept. It concludes with future looking platform requirements and how Adobe expects to take advantage of this technology at scale moving forward.