← all runs
Cohere-10M — 2026-07-07
r8i.4xlarge · peak recall 0.9994 @ 18 QPS
QPS at recall
| Recall | mkt-n40000-cctrue-fs16-soar-cfs |
|---|
| 0.90 | 709 |
| 0.95 | 355 |
| 0.99 | 94 |
Latency at recall
| mkt-n40000-cctrue-fs16-soar-cfs |
|---|
| @ recall 0.90 | |
| p50 | 1.37 ms |
| p95 | 1.86 ms |
| p99 | 2.08 ms |
| @ recall 0.95 | |
| p50 | 2.75 ms |
| p95 | 3.51 ms |
| p99 | 3.83 ms |
| @ recall 0.99 | |
| p50 | 10.44 ms |
| p95 | 12.64 ms |
| p99 | 13.30 ms |
All values are taken at exactly the stated recall: each configuration's own sweep is log-interpolated between its two surrounding measured points (measured values are used verbatim on an exact hit) and the best configuration at that recall is shown — values are never interpolated across different configurations. “—” means no single sweep spans that recall; nothing is extrapolated. Recall is distance-based (ann-benchmarks, tie-aware). Insert/update/delete workloads are not measured.
Build comparison
| mkt-n40000-cctrue-fs16-soar-cfs |
|---|
| Build time | 3575 s |
| Index size | 2.1 GB |
Plots
Pareto svg ·
Distance-recall svg ·
Per-config svg ·
Build times svg ·
Index sizes svgHost
| Instance type | r8i.4xlarge |
| Hostname | ip-172-31-72-93 |
| CPU | Intel(R) Xeon(R) 6975P-C (16 vCPU) |
| Memory | 124 GB |
| OS | Ubuntu 26.04 LTS |
| Kernel | 7.0.0-1008-aws |
Versions
| rekall | 0.1.0 (8aa6a65 / main) |
| PostgreSQL | PostgreSQL 18.4 (Ubuntu 18.4-1.pgdg26.04+1) on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 15.2.0-16ubuntu1) 15.2.0, 64-bit |
| meerkat | 0.1.0 (4a39b1c) |
| pg_prewarm | 1.2 |
| vector | 0.8.4 |
Search
Index builds
| Index | Type | Build time | Size | Workers |
|---|
| mkt-n40000-cctrue-fs16-soar-cfs | meerkat | 3574.6s | 2.1 GB | 8 |
Index parameters
mkt-n40000-cctrue-fs16-soar-cfs: {"boundary_epsilon":0.05,"centroid_compression":true,"centroid_fastscan":true,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":40000,"rerank_vectors":null,"soar_lambda":1.0}
PostgreSQL settings
effective_cache_size = 96 GB
effective_io_concurrency = 256
huge_pages = try
maintenance_work_mem = 51 GB
max_connections = 100
max_parallel_maintenance_workers = 8
max_parallel_workers = 16
random_page_cost = 1.1
shared_buffers = 32 GB
wal_level = replica
work_mem = 1 GB
