ATC Experiments
Match the evidence to the claim. ATC is the systems community's implementation-and-measurement
venue: reviewers read for measured behavior on a real system, not asymptotics or accuracy on a
dataset. In round two, 3-4 reviewers close to your subarea will open the artifact and probe whether
the numbers are end-to-end, fair, and honest about cost. Design the evaluation so their first three
objections are already answered.
Match evidence to claim shape
| If your claim is... |
The evidence ATC expects |
| "Faster / lower latency" |
End-to-end latency including tails (p99/p999) and throughput at a matched operating point, on a described testbed |
| "Lower overhead / cheaper" |
The resource cost (CPU, memory, writes, energy) measured, at matched function — not just the headline win |
| "Scales" |
Measurements across a real range of load/nodes/cores with the scaling curve and where it bends |
| "More reliable / correct" |
Fault-injection or crash/recovery experiments, not just steady-state runs |
| "Useful in practice" (experience) |
Production-derived workloads and lessons; what broke and what generalizes |
Real testbeds and workloads
- Describe the testbed so results are reproducible: CPU/NIC/SSD models, core counts, memory,
kernel/OS versions, network topology, and any co-location. A result without its testbed is not a
systems result.
- Use realistic workloads. Production-derived traces, standard benchmarks, or documented
generators beat hand-picked inputs. State how the workload was obtained and why it is
representative; if it is synthetic, justify the parameters.
- Warm-up and steady state. Say how you handled cold start, warm-up windows, and measurement
duration — systems reviewers know where transient effects hide.
Fair baselines
- Compare against the strongest reasonable alternative, configured well (a strawman baseline is
caught immediately). If you tuned your system, tune the baseline.
- Compare at a matched cost or operating point: same memory budget, same flash-write budget,
same load. An unmatched comparison is the classic systems-reviewer objection.
- If no baseline exists, say so and use the unmodified system or an ablation of your own design
as the reference.
End-to-end plus microbenchmarks
ATC reviewers want both:
- End-to-end results show the contribution matters for the whole system under a real workload.
- Microbenchmarks isolate the mechanism, attributing the win (or cost) to your design rather
than to unrelated system effects. A paper with only end-to-end numbers cannot explain why; one
with only microbenchmarks cannot show it matters.
Tails, variance, and honest reporting
- Report tail latency (p99, often p999), not just means — the tail is where systems pain lives.
- Report variance across repeated runs (multiple trials, min/max or CIs). A single run is a data
point, not a result.
- Report the cost beside the gain, at the matched operating point (see
atc-writing-style). A
win with an unstated cost reads as a hidden weakness.
- State negative or neutral regions honestly — "where the working set fits, our policy neither helps
nor hurts" builds more trust than a uniformly rosy curve.
Provenance you cannot reconstruct later
Pin these at collection time — they cannot be recovered at the deadline (see atc-reproducibility):
[Hardware] CPU/NIC/SSD models, core/memory counts, firmware where it matters
[Software] kernel/OS versions, library and compiler versions, config flags
[Workload] trace source and date, generator version and seeds, request mix
[Method] warm-up window, measurement duration, number of runs, aggregation
[Code] commit SHAs for the system and every baseline
Special cases
- Concurrency/nondeterminism: report the distribution and the scheduling/affinity settings, not
a lucky run.
- Energy/power claims: name the measurement instrument and boundary (wall vs. component).
- Security/isolation claims: state the threat model and what the measurement does and does not
cover.
- Experience papers: the "evaluation" is the deployment itself — scale, duration, incidents, and
transferable lessons; ATC's Deployed Systems lane values this even without a new mechanism.
Output format
[Claim -> evidence] each claim mapped to the experiment that supports it; gaps flagged
[Testbed] hardware/software/workload described enough to reproduce? yes/no
[Baselines] strongest alternative, well-configured, at a matched operating point? yes/no
[Depth] end-to-end AND microbenchmarks present? tails + variance reported?
[Honesty] costs reported beside gains? neutral/negative regions stated?
[Provenance] hardware/software/workload/method/code pinned at collection time? yes/no
1---2name: atc-experiments3description: Use when designing or auditing the evaluation of an ATC (ACM SIGOPS Annual Technical Conference, formerly USENIX ATC) systems paper — matching evidence to the claim with real testbeds, fair baselines, end-to-end plus microbenchmark results, tail-latency and variance reporting, workload realism, and honest cost accounting.4---56# ATC Experiments78Match the evidence to the claim. ATC is the systems community's **implementation-and-measurement**9venue: reviewers read for measured behavior on a real system, not asymptotics or accuracy on a10dataset. In round two, 3-4 reviewers close to your subarea will open the artifact and probe whether11the numbers are end-to-end, fair, and honest about cost. Design the evaluation so their first three12objections are already answered.1314## Match evidence to claim shape1516| If your claim is... | The evidence ATC expects |17|---|---|18| "Faster / lower latency" | End-to-end latency **including tails (p99/p999)** and throughput at a **matched** operating point, on a described testbed |19| "Lower overhead / cheaper" | The resource cost (CPU, memory, writes, energy) measured, at matched function — not just the headline win |20| "Scales" | Measurements across a real range of load/nodes/cores with the scaling curve and where it bends |21| "More reliable / correct" | Fault-injection or crash/recovery experiments, not just steady-state runs |22| "Useful in practice" (experience) | Production-derived workloads and lessons; what broke and what generalizes |2324## Real testbeds and workloads2526- **Describe the testbed** so results are reproducible: CPU/NIC/SSD models, core counts, memory,27 kernel/OS versions, network topology, and any co-location. A result without its testbed is not a28 systems result.29- **Use realistic workloads.** Production-derived traces, standard benchmarks, or documented30 generators beat hand-picked inputs. State how the workload was obtained and why it is31 representative; if it is synthetic, justify the parameters.32- **Warm-up and steady state.** Say how you handled cold start, warm-up windows, and measurement33 duration — systems reviewers know where transient effects hide.3435## Fair baselines3637- Compare against the **strongest reasonable** alternative, configured well (a strawman baseline is38 caught immediately). If you tuned your system, tune the baseline.39- Compare at a **matched cost or operating point**: same memory budget, same flash-write budget,40 same load. An unmatched comparison is the classic systems-reviewer objection.41- If no baseline exists, say so and use the **unmodified system** or an ablation of your own design42 as the reference.4344## End-to-end plus microbenchmarks4546ATC reviewers want both:4748- **End-to-end** results show the contribution matters for the whole system under a real workload.49- **Microbenchmarks** isolate the mechanism, attributing the win (or cost) to your design rather50 than to unrelated system effects. A paper with only end-to-end numbers cannot explain *why*; one51 with only microbenchmarks cannot show it *matters*.5253## Tails, variance, and honest reporting5455- Report **tail latency** (p99, often p999), not just means — the tail is where systems pain lives.56- Report **variance across repeated runs** (multiple trials, min/max or CIs). A single run is a data57 point, not a result.58- **Report the cost beside the gain**, at the matched operating point (see `atc-writing-style`). A59 win with an unstated cost reads as a hidden weakness.60- State negative or neutral regions honestly — "where the working set fits, our policy neither helps61 nor hurts" builds more trust than a uniformly rosy curve.6263## Provenance you cannot reconstruct later6465Pin these at collection time — they cannot be recovered at the deadline (see `atc-reproducibility`):6667```text68[Hardware] CPU/NIC/SSD models, core/memory counts, firmware where it matters69[Software] kernel/OS versions, library and compiler versions, config flags70[Workload] trace source and date, generator version and seeds, request mix71[Method] warm-up window, measurement duration, number of runs, aggregation72[Code] commit SHAs for the system and every baseline73```7475## Special cases7677- **Concurrency/nondeterminism:** report the distribution and the scheduling/affinity settings, not78 a lucky run.79- **Energy/power claims:** name the measurement instrument and boundary (wall vs. component).80- **Security/isolation claims:** state the threat model and what the measurement does and does not81 cover.82- **Experience papers:** the "evaluation" is the deployment itself — scale, duration, incidents, and83 transferable lessons; ATC's Deployed Systems lane values this even without a new mechanism.8485## Output format8687```text88[Claim -> evidence] each claim mapped to the experiment that supports it; gaps flagged89[Testbed] hardware/software/workload described enough to reproduce? yes/no90[Baselines] strongest alternative, well-configured, at a matched operating point? yes/no91[Depth] end-to-end AND microbenchmarks present? tails + variance reported?92[Honesty] costs reported beside gains? neutral/negative regions stated?93[Provenance] hardware/software/workload/method/code pinned at collection time? yes/no94```
brycewang-stanford