You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Run the Phase 2 benchmark campaign — ingest and reads at 1 s blocks #873
Run the full-history benchmarks against the three Phase 2 synthetic datasets at Phase-2 cadence (1 s blocks), and publish per-profile latency numbers against the Phase-2 targets. All three datasets are complete 20,000-ledger sets (#867).
Phase 2 profile and targets
SAC transfer
OZ token transfer
Soroswap swap
Traffic
5k TPS
4k TPS
1.5k TPS
Tx per ledger (TPL)
5000
4000
1500
Block time 1 s (applied at replay — a dataset bakes in only tx model + TPL), retention 6 months.
Targets this run measures against:
Ingest: Phase 2 defines no per-slice target for the meta-available-to-client path; the phase budget is 2.5 s end to end. The harness measures the ingest slice of that path; results are labeled as the ingest slice, never as the full path, and reported as a full distribution (composition with the read/stream slice is tracked separately).
Reads: GetTransaction / GetEvents / GetLedgers / GetTransactions meet their SLAs at the Phase-2 TPL and block time.
Inputs
Set 2026-07-18-apply-load-20k (generated under #867):
Dataset
TPL
Ledgers
Note
sac-5000
5000
20,000
custom_token-4000
4000
20,000
shared with the Phase-1 campaign
soroswap-1500
1500
20,000
shared with the Phase-1 campaign
Report
Per profile: p50 / p90 / p99 / max for the ingest slice at 1 s cadence, and read-endpoint latencies vs their SLAs. Use three-state verdicts: met / not met / not measured at this load. Where no per-slice target exists, report the distribution and mark the verdict not measured against a slice target.
Done when
All three Phase-2 datasets have published campaign results at 1 s cadence, with a summary table against the Phase-2 targets and explicit labeling of what was and was not measured.
Activity
marwen-abid commentedon Jul 24, 2026
Phase 2 (1 s blocks, p99 ≤ 400 ms):