I’m Han Im, a distributed storage engineer who measures down to the node I/O stack.

What I work on

  • Distributed storage — Ceph RGW → RADOS. Owning features upstream, not just fixing bugs.
  • Node I/O stack — io_uring, blk-mq, NVMe, network stack. Instrumented with eBPF; I dig only into the layer that is actually in the way.
  • Kernel, only when the measurement hits a kernel limit — liburing, io_uring, libceph/rbd.

How the measurements work

Each measurement has a question, a prediction written before measuring, the result, the raw data, and a way to rerun it. Upstream work is tracked on Upstream.

Production

I run 8PB of Ceph across 5 clusters (RBD, RGW, CephFS) plus IBM Storage Scale and VAST, on a GPU/AI cloud with QEMU/KVM and a RoCEv2 fabric.

Elsewhere

Organizing team at Ubuntu Korea Community. GitHub · LinkedIn