Week 9 Bench: Sharing a tenant
Working out, from the outside, whether two processes are colocated on the same physical machine
19 April 2027
Before the Bench#
Read the placement sections of Ristenpart et al., 2009 — the internal-IP map and the round-trip-time check, not the extraction results — before arriving. This Bench works from supplied data rather than a live cloud account.
In the Bench#
Using supplied round-trip-time and shared-cache-contention data from pairs of cloud instances, decide which pairs were probably colocated on the same physical host and which were not. The exercise sits one layer up from week 6’s cache measurement: there the question was which addresses a neighbour touched, here it is whether a neighbour exists on the same hardware at all, which is the precondition every cross-tenant channel in this course’s second half depends on.
Afterwards#
Note which signal you trusted most. Providers have spent a decade making exactly that signal noisier — worth knowing which one you would reach for first, and which one is likely already defended against.
The spec
What you bring to the session. Written so a reader can tell whether it has been met without asking you.
- you can list at least two observable signals that suggest colocation on shared cloud hardware
- you can distinguish, in supplied data, a colocation signal from ordinary network noise
- you can state what a cloud provider changed to make the technique you used harder, historically
Teaching team
- Marisol Quaye — convenor