The POSIX programming model has provided a long-standing abstraction for I/O, but contemporary systems increasingly expose interfaces that deviate from this model in pursuit of performance, scalability, and flexibility. This talk traces the evolution of shared-memory and asynchronous I/O mechanisms, beginning with early hypervisor-driven approaches such as Xen’s virtual device drivers, and continuing through to modern system interfaces including Linux’s io_uring and macOS’s libdispatch. These developments illustrate the growing fragmentation of the I/O landscape, but also highlight opportunities for unifying abstractions that balance expressivity with ergonomics for these highly concurrent interfaces.

As a case study, I will discuss how Docker integrates these heterogeneous interfaces into end-to-end data paths. Finally, I will present ongoing work in the OCaml 5 ecosystem, where effect-based concurrency abstractions, as implemented in the Eio library, offer a principled way of reconciling post-POSIX I/O mechanisms with high-level language runtimes. My perspective aims to inform discussions on virtual machine and runtime design, with implications for concurrency models, performance engineering, and cross-platform systems support.

Anil Madhavapeddy is the Professor of Planetary Computing at the University of Cambridge Computer Laboratory. He co-directs the Centre for Earth Observation and the Cambridge Centre for Carbon Credits (4C), with the aim of accelerating natural climate solutions that contribute towards ending deforestation and improving biodiversity globally. He has decades of experience with constructing Internet-scale systems, and has contributed to open-source projects such as OCaml, Docker, Xen, and OpenBSD, with users ranging from cloud computing providers to financial institutions to governments worldwide.