Job frameworks
AwaObanpg-bossProcrastinateRiver
A job is claimed by one worker, retried with backoff and completed or failed individually.
Benchmark results · 2026-11-11 to 2026-11-15
Every system ran the same scenarios, one after another, against a fresh PostgreSQL 18.6 cluster capped at 4 CPUs, on one host, in November 2026. In saturation, the Postgres restart and many queues, every system had a pool of workers + 8 connections per process. In the other scenarios Awa and Absurd had that pool and the other systems their adapter's default (4 connections for Procrastinate, 10 for pg-boss). Absolute numbers depend on this host; the comparisons within a section are the point.
Every adapter, scenario and raw summary is in the repository, and every run can be repeated on your own hardware. Corrections from the maintainers of any system are welcome.
| System | Library | Version | Contract |
|---|---|---|---|
| Absurd | absurd-sdk | 0.5.1 | Durable workflow |
| Awa | awa | 0.6.17+git.1a53fd9e | Job framework |
| Oban | oban | 2.24.2 | Job framework |
| pg-boss | pg-boss | 12.38.2 | Job framework |
| PGMQ | pgmq extension | v1.13.1 | Visibility queue |
| PgQue | pgque | v0.3.0-alpha.1-30-g8e3064c | Event queue |
| Procrastinate | procrastinate | 3.11.0 | Job framework |
| River | github.com/riverqueue/river | v0.51.2 | Job framework |
AwaObanpg-bossProcrastinateRiver
A job is claimed by one worker, retried with backoff and completed or failed individually.
Absurd
Tasks run as checkpointed steps that replay after an interruption; driven here as one-step jobs through its worker.
PGMQ
A message becomes invisible when read and reappears unless deleted; no per-job retry policy, scheduling or dead-letter handling beyond its archive.
PgQue
Consumers read batches cut by a ticker and acknowledge a whole batch in one statement. Driven here as competing consumers; batch acknowledgement is a lighter contract than per-job completion.
One headline measurement per scenario. The bar shows where each value falls between the best and worst in its row (longer is better); bars never compare across rows, and nothing is added up. Each row links to the full scenario.
| Scenario | Absurdworkflow | Awa | Oban | pg-boss | PGMQvisibility queue | PgQueevent batches | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Jobs/s, 64 workershigher is better | 578 jobs/s | 5,689 jobs/s | 3,993 jobs/s | 2,248 jobs/s | 5,379 jobs/s | 18.9K jobs/s | 459 jobs/s | 1,267 jobs/s |
| Pickup p99 at 800 jobs/slower is better | 1.99 min | 19 ms | 20 ms | 410 ms | 7 ms | 107 ms | 47.1 s | 43.7 s |
| Idle transactions/slower is better | 23.5 xacts/s | 60.1 xacts/s | 16 xacts/s | 21.9 xacts/s | 160 xacts/s | 40.6 xacts/s | 65.5 xacts/s | 24 xacts/s |
| Jobs/s after a Postgres restarthigher is better | 802 jobs/s | 802 jobs/s | 802 jobs/s | 668 jobs/s | 399 jobs/s | 805 jobs/s | 775 jobs/s | 634 jobs/s |
| Neighbour p99, queue at 4×lower is better | 14 ms | 6 ms | 12 ms | 10 ms | 18 ms | 2 ms | 5 ms | 7 ms |
| Dead catalog tuples, stalled CDClower is better | 22 tuples | 11.4K tuples | 115 tuples | 139 tuples | 20 tuples | 114 tuples | 25 tuples | 214 tuples |
| Scheduled herd lateness p99lower is better | 2.28 min | 12 s | 20.5 s | 1.29 min | 1.94 s | – | 41 s | 1.04 min |
| Good jobs/s in a failure stormhigher is better | 139 jobs/s | 138 jobs/s | 138 jobs/s | 132 jobs/s | 138 jobs/s | 138 jobs/s | 138 jobs/s | 138 jobs/s |
| Pickup p99 over 200 queueslower is better | 255 ms | 85 ms | 13 ms | 5 ms | 3 ms | 105 ms | 11 ms | 42 ms |
| Database after an 8 GiB drainlower is better | 8 GB | 21.3 MB | 8 GB | 8.35 GB | 16 GB | 8.33 GB | 7.35 GB | 8 GB |
| WAL/s with 40 s jobslower is better | 12.3 KB/s | 27.9 KB/s | 11.2 KB/s | 6.52 KB/s | 4.64 KB/s | 6.26 KB/s | 5.95 KB/s | 11.5 KB/s |
| Backlog drain ratehigher is better | – | 5,154 jobs/s | 4,206 jobs/s | 2,510 jobs/s | 14.5K jobs/s | 14.3K jobs/s | 579 jobs/s | 1,226 jobs/s |
| 1M abandoned jobs, all done againlower is better | 28.5 min | 20.5 min | not done in 30 min | 14 min | 5.88 min | 16.3 s | 32 min | – |
How many 1 ms jobs per second does one worker process complete as it gets more workers?
Setup. Out of the box: the producer keeps 4,000 jobs queued; one worker process with 4, 16, 64 or 128 workers and a pool of workers + 8 connections; 30 s warmup, then 150 s measured. Tuned: each system at the settings its own documentation points to, within one budget for all: at most 4 processes, 128 handlers and 200 connections, the same 1 ms jobs, measured at one queue and at its recommended number of queues.
| System | 4 workers |
|---|---|
| PGMQ | 3,494 jobs/s |
| PgQue | 3,215 jobs/s |
| pg-boss | 512 jobs/s |
| Awa | 488 jobs/s |
| Procrastinate | 474 jobs/s |
| Oban | 385 jobs/s |
| Absurd | 274 jobs/s |
| River | 79.2 jobs/s |
| System | 16 workers |
|---|---|
| PGMQ | 12K jobs/s |
| PgQue | 9,751 jobs/s |
| pg-boss | 2,048 jobs/s |
| Awa | 1,717 jobs/s |
| Oban | 1,312 jobs/s |
| Procrastinate | 454 jobs/s |
| Absurd | 440 jobs/s |
| River | 317 jobs/s |
| System | 64 workers |
|---|---|
| PgQue | 18.9K jobs/s |
| Awa | 5,689 jobs/s |
| PGMQ | 5,379 jobs/s |
| Oban | 3,993 jobs/s |
| pg-boss | 2,248 jobs/s |
| River | 1,267 jobs/s |
| Absurd | 578 jobs/s |
| Procrastinate | 459 jobs/s |
| System | 128 workers |
|---|---|
| PgQue | 23.3K jobs/s |
| Awa | 8,573 jobs/s |
| Oban | 5,418 jobs/s |
| PGMQ | 3,678 jobs/s |
| River | 2,509 jobs/s |
| pg-boss | 2,240 jobs/s |
| Absurd | 614 jobs/s |
| Procrastinate | 451 jobs/s |
| System | Best measured: completed jobs/s |
|---|---|
| PgQue | 47.3K jobs/s (tuned, 1 queue) |
| PGMQ | 12K jobs/s (16 workers, 1 process) |
| River | 10.9K jobs/s (tuned, 4 queues) |
| Awa | 10.8K jobs/s (tuned, 4 queues) |
| Oban | 5,673 jobs/s (tuned, 1 queue) |
| pg-boss | 3,495 jobs/s (tuned, 1 queue) |
| Absurd | 1,625 jobs/s (tuned, 4 queues) |
| Procrastinate | 1,472 jobs/s (tuned, 1 queue) |
| System | Out of the box, 128 workers | Best measured |
|---|---|---|
| PgQue | 23.3K jobs/s | 47.3K jobs/s |
| PGMQ | 3,678 jobs/s | 12K jobs/s |
| River | 2,509 jobs/s | 10.9K jobs/s |
| Awa | 8,573 jobs/s | 10.8K jobs/s |
| Oban | 5,418 jobs/s | 5,673 jobs/s |
| pg-boss | 2,240 jobs/s | 3,495 jobs/s |
| Absurd | 614 jobs/s | 1,625 jobs/s |
| Procrastinate | 451 jobs/s | 1,472 jobs/s |
| System | Tuned, 1 queue | Tuned, 4 queues |
|---|---|---|
| River | 8,113 jobs/s | 10.9K jobs/s |
| Awa | 8,550 jobs/s | 10.8K jobs/s |
| Absurd | 1,382 jobs/s | 1,625 jobs/s |
| Absurd | 4 × 32 handlers; 1 or 4 queues; producer batches of 256. |
|---|---|
| Awa | 1 × 128 handlers; 1 or 4 queues; 64 connections; lease rotation 1 s (the default); producer batches of 1,000. |
| Oban | 4 × 32 handlers; 1 queue; 40 connections each; dispatch_cooldown 1 ms (default 5 ms); producer batches of 1,000. |
| pg-boss | 4 × 32 handlers; 1 queue; pool of 40 each; batchSize 32, burstWhenBatchFull, LISTEN/NOTIFY; producer batches of 1,000. |
| PGMQ | 4 × 32 consumers; 1 or 16 queues, every consumer sweeping all queues; read batches of 32; ack with pgmq.delete; producer batches of 1,000. |
| PgQue | 4 processes × 4 cooperative subconsumers, 32 handlers each; 1 queue; ticker_loop every 10 ms; producer batches of 1,000. |
| Procrastinate | 4 × 32 handlers; 1 queue; pool of 33 each (concurrency + 1, as its docs advise); producer batches of 500. |
| River | 1 × 128 handlers; 1 or 4 queues; 32 connections; FetchCooldown 2 ms and FetchPollInterval 20 ms (River's own benchmark settings); producer batches of 1,000. |
Log scale.
| System | Peak dead tuples at 64 workers |
|---|---|
| Awa | 585 tuples |
| PgQue | 1,138 tuples |
| Absurd | 34K tuples |
| Procrastinate | 54.6K tuples |
| River | 148K tuples |
| pg-boss | 464K tuples |
| Oban | 558K tuples |
| PGMQ | 867K tuples |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 274 jobs/s | 488 jobs/s | 385 jobs/s | 512 jobs/s | 3,494 jobs/s | 3,215 jobs/s | 474 jobs/s | 79.2 jobs/s |
| Peak dead tuples (lower is better) | 239 tuples | 355 tuples | 46.2K tuples | 58.3K tuples | 448K tuples | 438 tuples | 49.7K tuples | 9,500 tuples |
| WAL per job (lower is better) | 2.25 KB | 2.37 KB | 4.3 KB | 2.55 KB | 1.7 KB | 428 B | 2.47 KB | 3.7 KB |
| Postgres CPU (lower is better) | 0.82 cores | 1.34 cores | 0.22 cores | 0.48 cores | 0.29 cores | 0.06 cores | 0.78 cores | 0.05 cores |
| Worker CPU (lower is better) | 0.13 cores | 0.15 cores | 0.36 cores | 0.08 cores | 0.13 cores | 0.14 cores | 0.7 cores | 0.02 cores |
| Postgres private memory, peak (lower is better) | 18.6 MB | 297 MB | 45.8 MB | 25 MB | 27.7 MB | 28.1 MB | 29.3 MB | 24.5 MB |
| Worker memory, peak (lower is better) | 44.8 MB | 11.5 MB | 2.13 GB | 65 MB | 43.3 MB | 55.8 MB | 40.4 MB | 10.3 MB |
| Database load (active sessions) (lower is better) | 0.97 sessions | 1.47 sessions | 0.68 sessions | 0.45 sessions | 0.51 sessions | 0.047 sessions | 1.86 sessions | 0.067 sessions |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 440 jobs/s | 1,717 jobs/s | 1,312 jobs/s | 2,048 jobs/s | 12K jobs/s | 9,751 jobs/s | 454 jobs/s | 317 jobs/s |
| Peak dead tuples (lower is better) | 11.6K tuples | 353 tuples | 154K tuples | 243K tuples | 2.14M tuples | 725 tuples | 54.2K tuples | 37.6K tuples |
| WAL per job (lower is better) | 2.35 KB | 1.7 KB | 4.75 KB | 2.77 KB | 2.05 KB | 432 B | 2.48 KB | 3.77 KB |
| Postgres CPU (lower is better) | 0.74 cores | 1.44 cores | 0.58 cores | 1.95 cores | 1.73 cores | 0.12 cores | 0.81 cores | 0.06 cores |
| Worker CPU (lower is better) | 0.2 cores | 0.2 cores | 1.36 cores | 0.32 cores | 0.32 cores | 0.32 cores | 0.7 cores | 0.03 cores |
| Postgres private memory, peak (lower is better) | 19.4 MB | 345 MB | 76.4 MB | 47.9 MB | 67.5 MB | 30.6 MB | 29.6 MB | 21.7 MB |
| Worker memory, peak (lower is better) | 44.8 MB | 21 MB | 2.16 GB | 92.1 MB | 45.3 MB | 55.6 MB | 40.4 MB | 14.3 MB |
| Database load (active sessions) (lower is better) | 0.87 sessions | 1.47 sessions | 1 sessions | 2.11 sessions | 3.05 sessions | 0.16 sessions | 1.75 sessions | 0.053 sessions |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 578 jobs/s | 5,689 jobs/s | 3,993 jobs/s | 2,248 jobs/s | 5,379 jobs/s | 18.9K jobs/s | 459 jobs/s | 1,267 jobs/s |
| Peak dead tuples (lower is better) | 34K tuples | 585 tuples | 558K tuples | 464K tuples | 867K tuples | 1,138 tuples | 54.6K tuples | 148K tuples |
| WAL per job (lower is better) | 2.47 KB | 1.52 KB | 5.4 KB | 3.13 KB | 2.01 KB | 436 B | 2.47 KB | 4.05 KB |
| Postgres CPU (lower is better) | 0.66 cores | 1.57 cores | 1.55 cores | 2.89 cores | 3.94 cores | 0.21 cores | 0.82 cores | 0.13 cores |
| Worker CPU (lower is better) | 0.25 cores | 0.37 cores | 4.13 cores | 0.35 cores | 0.27 cores | 0.62 cores | 0.7 cores | 0.09 cores |
| Postgres private memory, peak (lower is better) | 18.6 MB | 329 MB | 258 MB | 104 MB | 200 MB | 30.8 MB | 32.6 MB | 25 MB |
| Worker memory, peak (lower is better) | 45.1 MB | 63.7 MB | 2.2 GB | 201 MB | 50.2 MB | 55.6 MB | 40.4 MB | 16.2 MB |
| Database load (active sessions) (lower is better) | 0.79 sessions | 1.99 sessions | 11 sessions | 2.77 sessions | 41.6 sessions | 0.25 sessions | 1.96 sessions | 0.23 sessions |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 614 jobs/s | 8,573 jobs/s | 5,418 jobs/s | 2,240 jobs/s | 3,678 jobs/s | 23.3K jobs/s | 451 jobs/s | 2,509 jobs/s |
| Peak dead tuples (lower is better) | 32.5K tuples | 570 tuples | 911K tuples | 471K tuples | 595K tuples | 1,308 tuples | 52.6K tuples | 325K tuples |
| WAL per job (lower is better) | 2.52 KB | 1.51 KB | 5.58 KB | 3.15 KB | 1.97 KB | 434 B | 2.46 KB | 4.28 KB |
| Postgres CPU (lower is better) | 0.63 cores | 1.64 cores | 2.33 cores | 3 cores | 4 cores | 0.25 cores | 0.76 cores | 0.23 cores |
| Worker CPU (lower is better) | 0.26 cores | 0.5 cores | 6.7 cores | 0.36 cores | 0.27 cores | 0.76 cores | 0.67 cores | 0.18 cores |
| Postgres private memory, peak (lower is better) | 18.4 MB | 332 MB | 432 MB | 128 MB | 345 MB | 30.4 MB | 33.9 MB | 27.9 MB |
| Worker memory, peak (lower is better) | 45.5 MB | 69.3 MB | 2.23 GB | 188 MB | 52.4 MB | 55.6 MB | 40.4 MB | 17.9 MB |
| Database load (active sessions) (lower is better) | 0.81 sessions | 1.7 sessions | 12.9 sessions | 2.89 sessions | 73.1 sessions | 0.33 sessions | 2.06 sessions | 0.31 sessions |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 1,382 jobs/s | 8,550 jobs/s | 5,673 jobs/s | 3,495 jobs/s | 7,388 jobs/s | 47.3K jobs/s | 1,472 jobs/s | 8,113 jobs/s |
| Peak dead tuples (lower is better) | 92.2K tuples | 548 tuples | 1.05M tuples | 639K tuples | 1.55M tuples | 6,342 tuples | 172K tuples | 1.51M tuples |
| WAL per job (lower is better) | 2.64 KB | 1.54 KB | 6.45 KB | 3.79 KB | 1.35 KB | 443 B | 2.47 KB | 4.7 KB |
| Postgres CPU (lower is better) | 2.69 cores | 1.57 cores | 3.78 cores | 3.94 cores | 4 cores | 1.09 cores | 2.23 cores | 1.05 cores |
| Worker CPU (lower is better) | 0.64 cores | 0.51 cores | 6.2 cores | 0.77 cores | 0.39 cores | 1.9 cores | 2.02 cores | 0.52 cores |
| Postgres private memory, peak (lower is better) | 43.9 MB | 353 MB | 418 MB | 377 MB | 343 MB | 216 MB | 76.4 MB | 45.9 MB |
| Worker memory, peak (lower is better) | 173 MB | 71.8 MB | 8.65 GB | 340 MB | 177 MB | 260 MB | 162 MB | 19.9 MB |
| Database load (active sessions) (lower is better) | 3.86 sessions | 1.96 sessions | 23.3 sessions | 38.3 sessions | 104 sessions | 5.39 sessions | 5.39 sessions | 1.7 sessions |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 1,625 jobs/s | 10.8K jobs/s | – | – | – | – | – | 10.9K jobs/s |
| Peak dead tuples (lower is better) | 40.8K tuples | 708 tuples | – | – | – | – | – | 2.2M tuples |
| WAL per job (lower is better) | 3.21 KB | 1.62 KB | – | – | – | – | – | 5.25 KB |
| Postgres CPU (lower is better) | 3.99 cores | 3.96 cores | – | – | – | – | – | 2.16 cores |
| Worker CPU (lower is better) | 0.95 cores | 0.83 cores | – | – | – | – | – | 0.74 cores |
| Postgres private memory, peak (lower is better) | 103 MB | 456 MB | – | – | – | – | – | 55.1 MB |
| Worker memory, peak (lower is better) | 173 MB | 77 MB | – | – | – | – | – | 22.1 MB |
| Database load (active sessions) (lower is better) | 12.5 sessions | 5.01 sessions | – | – | – | – | – | 4.41 sessions |
Raw summaries: Absurd ↗Awa ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | – | – | – | – | 5,710 jobs/s | – | – | – |
| Peak dead tuples (lower is better) | – | – | – | – | 61.7K tuples | – | – | – |
| WAL per job (lower is better) | – | – | – | – | 1.33 KB | – | – | – |
| Postgres CPU (lower is better) | – | – | – | – | 4 cores | – | – | – |
| Worker CPU (lower is better) | – | – | – | – | 0.88 cores | – | – | – |
| Postgres private memory, peak (lower is better) | – | – | – | – | 382 MB | – | – | – |
| Worker memory, peak (lower is better) | – | – | – | – | 177 MB | – | – | – |
| Database load (active sessions) (lower is better) | – | – | – | – | 69.9 sessions | – | – | – |
Raw summaries: PGMQ ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 1,625 jobs/s | 10.8K jobs/s | 5,673 jobs/s | 3,495 jobs/s | 12K jobs/s | 47.3K jobs/s | 1,472 jobs/s | 10.9K jobs/s |
| Peak dead tuples (lower is better) | 40.8K tuples | 708 tuples | 1.05M tuples | 639K tuples | 2.14M tuples | 6,342 tuples | 172K tuples | 2.2M tuples |
| WAL per job (lower is better) | 3.21 KB | 1.62 KB | 6.45 KB | 3.79 KB | 2.05 KB | 443 B | 2.47 KB | 5.25 KB |
| Postgres CPU (lower is better) | 3.99 cores | 3.96 cores | 3.78 cores | 3.94 cores | 1.73 cores | 1.09 cores | 2.23 cores | 2.16 cores |
| Worker CPU (lower is better) | 0.95 cores | 0.83 cores | 6.2 cores | 0.77 cores | 0.32 cores | 1.9 cores | 2.02 cores | 0.74 cores |
| Postgres private memory, peak (lower is better) | 103 MB | 456 MB | 418 MB | 377 MB | 67.5 MB | 216 MB | 76.4 MB | 55.1 MB |
| Worker memory, peak (lower is better) | 173 MB | 77 MB | 8.65 GB | 340 MB | 45.3 MB | 260 MB | 162 MB | 22.1 MB |
| Database load (active sessions) (lower is better) | 12.5 sessions | 5.01 sessions | 23.3 sessions | 38.3 sessions | 3.05 sessions | 5.39 sessions | 5.39 sessions | 4.41 sessions |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
At a rate every system should sustain, how quickly does a job go from insert to completion?
Setup. 800 jobs/s offered, 32 workers, one process; 60 s warmup, then 300 s measured.
Log scale.
| System | Pickup latency, p99 |
|---|---|
| PGMQ | 7 ms |
| Awa | 19 ms |
| Oban | 20 ms |
| PgQue | 107 ms |
| pg-boss | 410 ms |
| River | 43.7 s (reached 634 jobs/s) |
| Procrastinate | 47.1 s (reached 442 jobs/s) |
| Absurd | 1.99 min (reached 223 jobs/s) |
| System | WAL written per job |
|---|---|
| PgQue | 457 B |
| PGMQ | 1.47 KB |
| Awa | 1.7 KB |
| pg-boss | 2.8 KB |
| Procrastinate | 3.1 KB |
| River | 4.57 KB |
| Oban | 4.98 KB |
| Absurd | 5.19 KB |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 223 jobs/s | 799 jobs/s | 800 jobs/s | 827 jobs/s | 803 jobs/s | 804 jobs/s | 442 jobs/s | 634 jobs/s |
| End-to-end p99 (lower is better) | 1.99 min | 21 ms | – | 426 ms | 14 ms | 108 ms | – | – |
| Claim p99 (lower is better) | 1.99 min | 19 ms | 20 ms | 410 ms | 7 ms | 107 ms | 47.1 s | 43.7 s |
| Queue depth (lower is better) | 95.7K jobs | 0 jobs | 9 jobs | 141 jobs | 1 jobs | 0 jobs | 28.9K jobs | 35K jobs |
| WAL per job (lower is better) | 5.19 KB | 1.7 KB | 4.98 KB | 2.8 KB | 1.47 KB | 457 B | 3.1 KB | 4.57 KB |
| Postgres CPU (lower is better) | 1.04 cores | 0.61 cores | 0.36 cores | 0.83 cores | 0.59 cores | 0.07 cores | 0.91 cores | 0.14 cores |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
With workers running and no jobs, how much does each system cost the database?
Setup. One process with 8 workers after a short warmup; 10 minutes with nothing enqueued.
| System | Transactions per second while idle |
|---|---|
| Oban | 16 xacts/s |
| pg-boss | 21.9 xacts/s |
| Absurd | 23.5 xacts/s |
| River | 24 xacts/s |
| PgQue | 40.6 xacts/s |
| Awa | 60.1 xacts/s |
| Procrastinate | 65.5 xacts/s |
| PGMQ | 160 xacts/s |
| System | WAL written while idle |
|---|---|
| PGMQ | 429 B/s |
| Procrastinate | 459 B/s |
| Absurd | 991 B/s |
| pg-boss | 1.77 KB/s |
| Awa | 3.91 KB/s |
| Oban | 4.19 KB/s |
| PgQue | 7.67 KB/s |
| River | 11.4 KB/s |
| System | Postgres private memory while idle |
|---|---|
| Absurd | 17.7 MB |
| River | 23.5 MB |
| pg-boss | 24.7 MB |
| PgQue | 29.1 MB |
| Procrastinate | 32.7 MB |
| PGMQ | 37.8 MB |
| Oban | 58.7 MB |
| Awa | 276 MB |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Transactions/s (lower is better) | 23.5 xacts/s | 60.1 xacts/s | 16 xacts/s | 21.9 xacts/s | 160 xacts/s | 40.6 xacts/s | 65.5 xacts/s | 24 xacts/s |
| WAL written (lower is better) | 991 B/s | 3.91 KB/s | 4.19 KB/s | 1.77 KB/s | 429 B/s | 7.67 KB/s | 459 B/s | 11.4 KB/s |
| Table rewrites (lower is better) | 0 per s | 0.37 per s | 0 per s | 0 per s | 0 per s | 0 per s | 0 per s | 0 per s |
| Postgres CPU (lower is better) | 0.05 cores | 0.05 cores | 0.02 cores | 0.03 cores | 0.06 cores | 0.03 cores | 0.07 cores | 0.02 cores |
| Postgres private memory (lower is better) | 17.7 MB | 276 MB | 58.7 MB | 24.7 MB | 37.8 MB | 29.1 MB | 32.7 MB | 23.5 MB |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
When the database restarts mid-run, does the system come back and work off what accumulated?
Setup. Two processes, 400 jobs/s each, 16 workers each. 90 s baseline; Postgres stopped for 45 s and started again; 120 s recovery.
| System | Completed jobs/s after the restart |
|---|---|
| PgQue | 805 jobs/s |
| Awa | 802 jobs/s |
| Absurd | 802 jobs/s |
| Oban | 802 jobs/s |
| Procrastinate | 775 jobs/s |
| pg-boss | 668 jobs/s |
| River | 634 jobs/s |
| PGMQ | 399 jobs/s |
Log scale.
| System | Queue depth after the restart |
|---|---|
| Oban | 0 jobs |
| PgQue | 0 jobs |
| PGMQ | 1 jobs |
| Procrastinate | 2 jobs |
| Awa | 4 jobs |
| Absurd | 6 jobs |
| pg-boss | 176 jobs |
| River | 37.7K jobs |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Before restart (higher is better) | 794 jobs/s | 798 jobs/s | 793 jobs/s | 668 jobs/s | 401 jobs/s | 801 jobs/s | 777 jobs/s | 634 jobs/s |
| After restart (higher is better) | 802 jobs/s | 802 jobs/s | 802 jobs/s | 668 jobs/s | 399 jobs/s | 805 jobs/s | 775 jobs/s | 634 jobs/s |
| Workers back (lower is better) | 5.02 s | 5 s | 5.01 s | 5.01 s | 5.01 s | 5.01 s | 5 s | 5 s |
| Backlog cleared (lower is better) | 10 s | 10 s | 5 s | – | 5.01 s | 5.01 s | 5 s | – |
| End-to-end p99 after restart (lower is better) | 1.2 s | 30 ms | – | 400 ms | 28 ms | 109 ms | – | – |
| Queue depth after restart (lower is better) | 6 jobs | 4 jobs | 0 jobs | 176 jobs | 1 jobs | 0 jobs | 2 jobs | 37.7K jobs |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
Does a busy queue slow down unrelated transactions on the same server?
Setup. pgbench runs a TPC-B-like workload at a fixed 300 TPS on a separate database throughout. The queue runs at 0, 400, 1,600, then 400 jobs/s.
| System | Queue idle | Queue at 1,600 jobs/s |
|---|---|---|
| PGMQ | 2 ms | 18 ms |
| Absurd | 2 ms | 14 ms |
| Oban | 4 ms | 12 ms |
| pg-boss | 2 ms | 10 ms |
| River | 2 ms | 7 ms |
| Awa | 2 ms | 6 ms |
| Procrastinate | 2 ms | 5 ms |
| PgQue | 2 ms | 2 ms |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Neighbour p99, queue idle (lower is better) | 2 ms | 2 ms | 4 ms | 2 ms | 2 ms | 2 ms | 2 ms | 2 ms |
| Neighbour p99, queue at 1× (lower is better) | 6 ms | 4 ms | 5 ms | 5 ms | 5 ms | 3 ms | 5 ms | 9 ms |
| Neighbour p99, queue at 4× (lower is better) | 14 ms | 6 ms | 12 ms | 10 ms | 18 ms | 2 ms | 5 ms | 7 ms |
| Neighbour TPS vs idle (higher is better) | 1× | 0.99× | 1× | 1.01× | 1× | 1× | 1× | 1× |
| Queue jobs/s at 4× (higher is better) | 80 jobs/s | 1,600 jobs/s | 1,391 jobs/s | 512 jobs/s | 1,607 jobs/s | 1,606 jobs/s | 427 jobs/s | 317 jobs/s |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
Does the queue coexist with CDC on the same database, including when the CDC consumer stops?
Setup. A FOR ALL TABLES publication, a pgoutput slot and a streaming consumer; 400 jobs/s. The consumer is frozen for 4 minutes, then resumes.
| System | Published tables without a replica identity |
|---|---|
| Absurd | 0 tables |
| Awa | 0 tables |
| Oban | 0 tables |
| pg-boss | 0 tables |
| Procrastinate | 0 tables |
| River | 0 tables |
| PGMQ | 2 tables |
| PgQue | 5 tables |
Log scale.
| System | Dead catalog tuples while the consumer was stalled |
|---|---|
| PGMQ | 20 tuples |
| Absurd | 22 tuples |
| Procrastinate | 25 tuples |
| PgQue | 114 tuples |
| Oban | 115 tuples |
| pg-boss | 139 tuples |
| River | 214 tuples |
| Awa | 11.4K tuples |
| System | WAL retained by the stalled slot |
|---|---|
| PgQue | 48.1 MB |
| PGMQ | 169 MB |
| Absurd | 190 MB |
| Awa | 201 MB |
| pg-boss | 255 MB |
| Procrastinate | 348 MB |
| River | 389 MB |
| Oban | 569 MB |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Tables without replica identity (lower is better) | 0 tables | 0 tables | 0 tables | 0 tables | 2 tables | 5 tables | 0 tables | 0 tables |
| WAL retained by the stalled slot (lower is better) | 190 MB | 201 MB | 569 MB | 255 MB | 169 MB | 48.1 MB | 348 MB | 389 MB |
| Dead catalog tuples during the stall (lower is better) | 22 tuples | 11.4K tuples | 115 tuples | 139 tuples | 20 tuples | 114 tuples | 25 tuples | 214 tuples |
| WAL per job while streaming (lower is better) | 3.23 KB | 2.01 KB | 5.02 KB | 3.25 KB | 1.79 KB | 490 B | 2.87 KB | 4.82 KB |
| Jobs/s while streaming (higher is better) | 402 jobs/s | 399 jobs/s | 400 jobs/s | 334 jobs/s | 402 jobs/s | 402 jobs/s | 395 jobs/s | 317 jobs/s |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
When 20,000 jobs come due at the same instant, how late do they start, and do immediate jobs suffer?
Setup. 20,000 jobs scheduled for one instant while a 200 jobs/s stream of immediate jobs keeps flowing; 16 workers.
Log scale.
| System | Herd lateness, p99 |
|---|---|
| PGMQ | 1.94 s |
| Awa | 12 s |
| Oban | 20.5 s |
| Procrastinate | 41 s |
| River | 1.04 min |
| pg-boss | 1.29 min |
| Absurd | 2.28 min |
| PgQue | – |
Log scale.
| System | Immediate jobs while the herd drains, claim p99 |
|---|---|
| Procrastinate | 9 ms |
| Oban | 13 ms |
| Awa | 19 ms |
| PGMQ | 43 ms |
| PgQue | 107 ms |
| River | 9.21 s |
| pg-boss | 53.1 s |
| Absurd | 1.89 min |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Herd lateness p50 (lower is better) | 1.61 min | 6.33 s | 11.7 s | 39.3 s | 958 ms | – | 20.7 s | 31.6 s |
| Herd lateness p99 (lower is better) | 2.28 min | 12 s | 20.5 s | 1.29 min | 1.94 s | – | 41 s | 1.04 min |
| Herd drained (lower is better) | 2.29 min | 12.1 s | 20.7 s | 1.31 min | 1.95 s | – | 41.4 s | 1.05 min |
| Immediate jobs, claim p99 (lower is better) | 1.89 min | 19 ms | 13 ms | 53.1 s | 43 ms | 107 ms | 9 ms | 9.21 s |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
When a third of jobs start failing, does useful work keep flowing?
Setup. 200 jobs/s, 16 workers. For 6 minutes 30% of new jobs fail once and 1% always fail; at most 5 attempts; each system's own backoff applies.
| System | Good jobs completed per second during the storm |
|---|---|
| Absurd | 139 jobs/s |
| PgQue | 138 jobs/s |
| PGMQ | 138 jobs/s |
| River | 138 jobs/s |
| Awa | 138 jobs/s |
| Oban | 138 jobs/s |
| Procrastinate | 138 jobs/s |
| pg-boss | 132 jobs/s |
Log scale.
| System | Claim latency during the storm, p99 |
|---|---|
| Procrastinate | 11 ms |
| Absurd | 12 ms |
| Awa | 24 ms |
| PGMQ | 42 ms |
| Oban | 44 ms |
| River | 52 ms |
| PgQue | 107 ms |
| pg-boss | 8.33 s |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Before the storm (higher is better) | 201 jobs/s | 200 jobs/s | 200 jobs/s | 200 jobs/s | 201 jobs/s | 201 jobs/s | 200 jobs/s | 200 jobs/s |
| Good jobs/s during the storm (higher is better) | 139 jobs/s | 138 jobs/s | 138 jobs/s | 132 jobs/s | 138 jobs/s | 138 jobs/s | 138 jobs/s | 138 jobs/s |
| Claim p99 during the storm (lower is better) | 12 ms | 24 ms | 44 ms | 8.33 s | 42 ms | 107 ms | 11 ms | 52 ms |
| Retry backlog cleared (lower is better) | 3.15 s | 39.4 s | 1.58 min | 9.13 s | 2.08 min | 4.08 min | 4.37 s | – |
| Peak dead tuples (lower is better) | 7,651 tuples | 4,572 tuples | 34.6K tuples | 30K tuples | 26.9K tuples | 5,089 tuples | 31.6K tuples | 32.3K tuples |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
What does spreading the same load over 200 queues cost, and how does each system serve many queues?
Setup. One process. 200 jobs/s in total, round-robin over 1 queue and then over 200 queues (1 job/s per queue); 16 handlers and 24 connections requested; each system uses its own idiomatic way of serving many queues, the same in both runs; 4 minutes busy, then 3 minutes idle.
Log scale.
| System | 1 queue | 200 queues |
|---|---|---|
| Absurd | 10 ms | 255 ms |
| PgQue | 137 ms | 105 ms |
| Awa | 19 ms | 85 ms |
| River | 51 ms | 42 ms |
| Oban | 13 ms | 13 ms |
| Procrastinate | 6 ms | 11 ms |
| pg-boss | 5 ms | 5 ms |
| PGMQ | 4 ms | 3 ms |
| System | 1 queue | 200 queues |
|---|---|---|
| PgQue | 35.2 xacts/s | 1,498 xacts/s |
| Absurd | 8.37 xacts/s | 809 xacts/s |
| Awa | 60.1 xacts/s | 667 xacts/s |
| PGMQ | 36.3 xacts/s | 401 xacts/s |
| Procrastinate | 374 xacts/s | 374 xacts/s |
| River | 7.76 xacts/s | 203 xacts/s |
| Oban | 163 xacts/s | 168 xacts/s |
| pg-boss | 16.1 xacts/s | 48.9 xacts/s |
| System | 1 queue | 200 queues |
|---|---|---|
| Oban | 25 backends | 25 backends |
| PgQue | 22 backends | 22 backends |
| Awa | 11 backends | 20 backends |
| PGMQ | 19 backends | 19 backends |
| Absurd | 3 backends | 18 backends |
| River | 5 backends | 15 backends |
| pg-boss | 19 backends | 6 backends |
| Procrastinate | 6 backends | 6 backends |
| Absurd | One client per queue (200 handler slots), sharing 16 connections; 250 ms poll (the SDK default). |
|---|---|
| Awa | Weighted mode: one pool of 16 workers across every queue; one LISTEN connection. |
| Oban | One producer per queue with limit 1 (200 slots); idle queues don't poll. |
| pg-boss | One worker per queue (200); LISTEN/NOTIFY wakes idle workers. |
| PGMQ | 16 consumers sweeping their share of the queues; insert NOTIFY on one LISTEN connection; 500 ms fallback poll. |
| PgQue | Shared consumer over 16 connections; one LISTEN connection; ticker_loop every 50 ms. |
| Procrastinate | One worker over all 200 queues with concurrency 16; one LISTEN connection. |
| River | One producer per queue with MaxWorkers 1 (200); River's default 100 ms cooldown and 1 s poll; one notifier connection. |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 201 jobs/s | 200 jobs/s | 200 jobs/s | 200 jobs/s | 201 jobs/s | 200 jobs/s | 200 jobs/s | 200 jobs/s |
| Claim p99 (lower is better) | 10 ms | 19 ms | 13 ms | 5 ms | 4 ms | 137 ms | 6 ms | 51 ms |
| Postgres connections (lower is better) | 3 backends | 11 backends | 25 backends | 19 backends | 19 backends | 22 backends | 6 backends | 5 backends |
| LISTEN connections (lower is better) | 0 backends | 2 backends | 1 backends | 0 backends | 1 backends | 1 backends | 1 backends | 0 backends |
| Idle transactions/s (lower is better) | 8.37 xacts/s | 60.1 xacts/s | 163 xacts/s | 16.1 xacts/s | 36.3 xacts/s | 35.2 xacts/s | 374 xacts/s | 7.76 xacts/s |
| Idle Postgres CPU (lower is better) | 0.02 cores | 0.06 cores | 0.02 cores | 0.05 cores | 0.03 cores | 0.03 cores | 0.03 cores | 0.02 cores |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 201 jobs/s | 200 jobs/s | 200 jobs/s | 200 jobs/s | 201 jobs/s | 200 jobs/s | 200 jobs/s | 200 jobs/s |
| Claim p99 (lower is better) | 255 ms | 85 ms | 13 ms | 5 ms | 3 ms | 105 ms | 11 ms | 42 ms |
| Postgres connections (lower is better) | 18 backends | 20 backends | 25 backends | 6 backends | 19 backends | 22 backends | 6 backends | 15 backends |
| LISTEN connections (lower is better) | 0 backends | 2 backends | 1 backends | 0 backends | 1 backends | 1 backends | 1 backends | 0 backends |
| Idle transactions/s (lower is better) | 809 xacts/s | 667 xacts/s | 168 xacts/s | 48.9 xacts/s | 401 xacts/s | 1,498 xacts/s | 374 xacts/s | 203 xacts/s |
| Idle Postgres CPU (lower is better) | 1.68 cores | 0.24 cores | 0.02 cores | 0.98 cores | 0.11 cores | 1.05 cores | 0.04 cores | 0.05 cores |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
What happens with 64 KiB and 1 MiB arguments, and does the database shrink back afterwards?
Setup. Incompressible arguments. Saturation: one process, 64 workers, 2,000 jobs queued, 120 s. Backlog: about 8 GiB of 1 MiB jobs preloaded by four processes, then drained by 4 × 16 workers.
| System | 1 MiB args: MiB/s completed |
|---|---|
| River | 51.2 MiB/s |
| Awa | 51.2 MiB/s |
| Oban | 44.9 MiB/s |
| Absurd | 38.4 MiB/s |
| pg-boss | 25.6 MiB/s |
| PGMQ | 25.6 MiB/s |
| PgQue | 25.4 MiB/s |
| Procrastinate | 22 MiB/s |
| System | 64 KiB args: MiB/s completed |
|---|---|
| River | 49.9 MiB/s |
| Awa | 43.2 MiB/s |
| Oban | 41.5 MiB/s |
| pg-boss | 35.2 MiB/s |
| PGMQ | 26.5 MiB/s |
| PgQue | 23.9 MiB/s |
| Absurd | 16.8 MiB/s |
| Procrastinate | 12.4 MiB/s |
Log scale.
| System | Database size after draining the 1 MiB backlog |
|---|---|
| Awa | 21.3 MB |
| Procrastinate | 7.35 GB |
| River | 8 GB |
| Oban | 8 GB |
| Absurd | 8 GB |
| PgQue | 8.33 GB |
| pg-boss | 8.35 GB |
| PGMQ | 16 GB |
Log scale.
| System | Worker memory while draining, all four processes |
|---|---|
| Procrastinate | 216 MB |
| PgQue | 217 MB |
| Awa | 259 MB |
| Absurd | 366 MB |
| River | 379 MB |
| PGMQ | 998 MB |
| pg-boss | 5.32 GB |
| Oban | 8.71 GB |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 269 jobs/s | 691 jobs/s | 664 jobs/s | 563 jobs/s | 424 jobs/s | 383 jobs/s | 199 jobs/s | 798 jobs/s |
| WAL per job (lower is better) | 113 KB | 94.8 KB | 103 KB | 73.6 KB | 188 KB | 71 KB | 110 KB | 86.5 KB |
| Worker memory, peak (lower is better) | 59.8 MB | 88 MB | 2.26 GB | 366 MB | 233 MB | 98.8 MB | 40.6 MB | 52.9 MB |
| Database, peak (lower is better) | 3.31 GB | 1.69 GB | 9.92 GB | 6.37 GB | 24.7 GB | 3.9 GB | 1.89 GB | 12 GB |
| Database after the drain (lower is better) | 3.31 GB | 1.22 GB | 9.92 GB | 6.37 GB | 24.7 GB | 3.9 GB | 1.89 GB | 12 GB |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 38.4 jobs/s | 51.2 jobs/s | 44.9 jobs/s | 25.6 jobs/s | 25.6 jobs/s | 25.4 jobs/s | 22 jobs/s | 51.2 jobs/s |
| WAL per job (lower is better) | 1.27 MB | 1.42 MB | 1.2 MB | 1.49 MB | 3.62 MB | 1.12 MB | 1.96 MB | 1.24 MB |
| Worker memory, peak (lower is better) | 372 MB | 944 MB | 2.55 GB | 552 MB | 3.08 GB | 808 MB | 57.1 MB | 531 MB |
| Database, peak (lower is better) | 15.9 GB | 1.63 GB | 10.3 GB | 7.25 GB | 17 GB | 4.05 GB | 3.44 GB | 19.6 GB |
| Database after the drain (lower is better) | 15.9 GB | 823 MB | 10.3 GB | 7.25 GB | 17 GB | 4.05 GB | 3.44 GB | 19.6 GB |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Jobs preloaded (higher is better) | 7,684 jobs | 6,494 jobs | 7,675 jobs | 8,004 jobs | 7,684 jobs | 7,994 jobs | 7,056 jobs | 7,680 jobs |
| Drain rate (higher is better) | 243 jobs/s | 170 jobs/s | 258 jobs/s | 616 jobs/s | 45.2 jobs/s | 231 jobs/s | 184 jobs/s | 279 jobs/s |
| Worker memory while draining (lower is better) | 366 MB | 259 MB | 8.71 GB | 5.32 GB | 998 MB | 217 MB | 216 MB | 379 MB |
| Database, peak (lower is better) | 8 GB | 6.77 GB | 8 GB | 8.35 GB | 16 GB | 8.33 GB | 7.35 GB | 8 GB |
| Database after the drain (lower is better) | 8 GB | 21.3 MB | 8 GB | 8.35 GB | 16 GB | 8.33 GB | 7.35 GB | 8 GB |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
At a gentle rate with 64 KiB arguments, how much WAL is written and how much space is left behind?
Setup. 64 KiB incompressible arguments at 50 jobs/s for 5 minutes, then 3 minutes of draining.
Log scale.
| System | TOAST storage after draining |
|---|---|
| Awa | 625 KB |
| Procrastinate | 1.16 GB |
| Oban | 1.17 GB |
| River | 1.17 GB |
| Absurd | 1.17 GB |
| pg-boss | 1.17 GB |
| PGMQ | 1.18 GB |
| PgQue | 1.18 GB |
| System | WAL written per job |
|---|---|
| PgQue | 70.9 KB |
| Awa | 75.3 KB |
| Procrastinate | 80.7 KB |
| River | 81 KB |
| Oban | 83.1 KB |
| pg-boss | 84.2 KB |
| Absurd | 86.7 KB |
| PGMQ | 205 KB |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Completed jobs/s (higher is better) | 50 jobs/s | 50 jobs/s | 50 jobs/s | 50 jobs/s | 50 jobs/s | 50.1 jobs/s | 49.8 jobs/s | 50 jobs/s |
| WAL per job (lower is better) | 86.7 KB | 75.3 KB | 83.1 KB | 84.2 KB | 205 KB | 70.9 KB | 80.7 KB | 81 KB |
| TOAST after draining (lower is better) | 1.17 GB | 625 KB | 1.17 GB | 1.17 GB | 1.18 GB | 1.18 GB | 1.16 GB | 1.17 GB |
| Database after draining (lower is better) | 1.19 GB | 22.2 MB | 1.19 GB | 1.19 GB | 1.19 GB | 1.19 GB | 1.18 GB | 1.19 GB |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
What does it cost to keep jobs running for 40 seconds each?
Setup. Two processes × 100 workers; 40 s jobs; 150 held, then a steady stream for 6 minutes.
| System | WAL written per second at steady state |
|---|---|
| PGMQ | 4.64 KB/s |
| Procrastinate | 5.95 KB/s |
| PgQue | 6.26 KB/s |
| pg-boss | 6.52 KB/s |
| Oban | 11.2 KB/s |
| River | 11.5 KB/s |
| Absurd | 12.3 KB/s |
| Awa | 27.9 KB/s |
| System | Duplicate completions |
|---|---|
| PgQue | -842 jobs |
| pg-boss | 0 jobs |
| Procrastinate | 0.0021 jobs |
| Awa | 0.0057 jobs |
| Oban | 298 jobs |
| River | 360 jobs |
| PGMQ | 610 jobs |
| Absurd | 1,134 jobs |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| WAL written (lower is better) | 12.3 KB/s | 27.9 KB/s | 11.2 KB/s | 6.52 KB/s | 4.64 KB/s | 6.26 KB/s | 5.95 KB/s | 11.5 KB/s |
| Transactions/s (lower is better) | 14.4 xacts/s | 115 xacts/s | 15.9 xacts/s | 260 xacts/s | 244 xacts/s | 47 xacts/s | 103 xacts/s | 52.3 xacts/s |
| Rows updated/s (lower is better) | 17 rows/s | 24.2 rows/s | 5.51 rows/s | 4.82 rows/s | 5.89 rows/s | 0.72 rows/s | 4.47 rows/s | 6.13 rows/s |
| Duplicate completions (lower is better) | 1,134 jobs | 0.0057 jobs | 298 jobs | 0 jobs | 610 jobs | -842 jobs | 0.0021 jobs | 360 jobs |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
How fast does a 200,000-job backlog drain, and how big is the database afterwards?
Setup. 200,000 jobs preloaded behind a gate, then drained by 64 workers, then 4 minutes idle.
| System | Drain rate |
|---|---|
| PGMQ | 14.5K jobs/s |
| PgQue | 14.3K jobs/s |
| Awa | 5,154 jobs/s |
| Oban | 4,206 jobs/s |
| pg-boss | 2,510 jobs/s |
| River | 1,226 jobs/s |
| Procrastinate | 579 jobs/s |
| Absurd | – |
Log scale.
| System | Database size after draining |
|---|---|
| Awa | 21.5 MB |
| PgQue | 75.9 MB |
| Procrastinate | 92.1 MB |
| PGMQ | 97.3 MB |
| pg-boss | 205 MB |
| Absurd | 211 MB |
| River | 216 MB |
| Oban | 323 MB |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Drain rate (higher is better) | – | 5,154 jobs/s | 4,206 jobs/s | 2,510 jobs/s | 14.5K jobs/s | 14.3K jobs/s | 579 jobs/s | 1,226 jobs/s |
| Time to empty (lower is better) | – | 38.8 s | 47.3 s | 1.2 min | 13.8 s | 14 s | 3.97 min | 2.72 min |
| Database before (lower is better) | 7.83 MB | 20.7 MB | 7.81 MB | 8.43 MB | 7.86 MB | 8.7 MB | 7.81 MB | 7.94 MB |
| Database, peak (lower is better) | 211 MB | 173 MB | 323 MB | 205 MB | 179 MB | 75.9 MB | 92.1 MB | 216 MB |
| Database after (lower is better) | 211 MB | 21.5 MB | 323 MB | 205 MB | 97.3 MB | 75.9 MB | 92.1 MB | 216 MB |
| Back to baseline size (lower is better) | – | 1.22 min | – | – | – | – | – | – |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗
With a million jobs queued, how long does an operator wait to find one?
Setup. A million jobs sit in a queue no worker consumes, each tagged with a unique handle and one of 1,000 tenants, while 200 jobs/s of ordinary work runs on another queue. Each probe runs through the system's own API (or SQL where that is the interface) at 2 per second for 3 minutes, on its own connection.
Log scale.
| System | Get one job by id, p99 |
|---|---|
| River | 0.85 ms |
| PGMQ | 0.88 ms (SQL) |
| Oban | 1 ms |
| Absurd | 1 ms |
| Procrastinate | 2 ms |
| pg-boss | 2 ms |
| Awa | 103 ms |
| Awa 0.7 (unreleased) | 551 ms |
| PgQue | – (not supported) |
Log scale.
| System | Find one job by a unique field, p99 |
|---|---|
| Absurd | 80 ms (SQL) |
| PGMQ | 164 ms (SQL) |
| pg-boss | 205 ms |
| Oban | 206 ms |
| River | 290 ms |
| Awa 0.7 (unreleased) | 588 ms |
| Awa | 30 s (timed out (30 s cap)) |
| PgQue | – (not supported) |
| Procrastinate | – (not supported) |
Log scale.
| System | First page of one tenant's 1,000 jobs, p99 |
|---|---|
| Oban | 28 ms |
| River | 29 ms |
| PGMQ | 60 ms (SQL) |
| pg-boss | 70 ms |
| Absurd | 93 ms (SQL) |
| Awa 0.7 (unreleased) | 665 ms |
| Awa | 30 s (timed out (30 s cap)) |
| PgQue | – (not supported) |
| Procrastinate | – (not supported) |
Log scale.
| System | First page by kind and state, p99 |
|---|---|
| PGMQ | 2 ms (SQL) |
| pg-boss | 3 ms |
| River | 4 ms |
| Oban | 4 ms |
| Absurd | 126 ms (SQL) |
| Procrastinate | 433 ms |
| Awa 0.7 (unreleased) | 903 ms |
| Awa | 2.4 s |
| PgQue | – (not supported) |
Log scale.
| System | Count by state, p99 |
|---|---|
| PgQue | 0.94 ms |
| pg-boss | 1 ms |
| PGMQ | 48 ms |
| Absurd | 58 ms (SQL) |
| River | 138 ms (SQL) |
| Oban | 147 ms |
| Awa | 149 ms |
| Procrastinate | 655 ms |
| Awa 0.7 (unreleased) | 1.19 s |
| Absurd | get_by_id through the SDK (fetch_task_result); the filters and count are the SQL its absurdctl list-tasks builds, against its task tables. |
|---|---|
| Awa | 0.6: get_job, list_jobs (tag, kind, state, queue; tags are checked after loading, with no index), queue_overview for counts. 0.7: find_jobs with a per-partition GIN index on tags. |
| Oban | Oban.Job.query (id, args containment, worker, state, queue) through the Repo; GIN index on args. |
| pg-boss | findJobs by id or data within one queue; getQueue counts are cached by the monitor (refreshed every 60 s). |
| PGMQ | SQL on the queue table, which is pgmq's interface (by msg_id, message containment); pgmq.metrics for the count. |
| PgQue | Counts only (get_queue_info); events can't be looked up. |
| Procrastinate | list_jobs by id, queue, task and status, with no LIMIT and no filter on arguments; list_queues for counts. |
| River | JobGet and JobList (metadata containment, kind, state, queue) with a GIN index on metadata; counts by SQL (no count API). |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River | Awa 0.7 (unreleased) |
|---|---|---|---|---|---|---|---|---|---|
| Time to enqueue the haystack (lower is better) | 2.23 min | 34.6 s | 1.21 min | 49.9 s | 13.3 s | 11.5 s | 1.51 min | 29.8 s | 46.9 s |
| Get by id, p50 (lower is better) | 0.76 ms | 10 ms | 0.36 ms | 1 ms | 0.32 ms | – | 0.96 ms | 0.27 ms | 471 ms |
| Get by id, p99 (lower is better) | 1 ms | 103 ms | 1 ms | 2 ms | 0.88 ms | – | 2 ms | 0.85 ms | 551 ms |
| Get by id: API (higher is better) | 1 mode | 1 mode | 1 mode | 1 mode | 2 mode | 0 mode | 1 mode | 1 mode | 1 mode |
| Get by id: correct answers (higher is better) | 1 fraction | 1 fraction | 1 fraction | 1 fraction | 1 fraction | – | 1 fraction | 1 fraction | 1 fraction |
| Find one job by a unique field, p50 (lower is better) | 73 ms | 30 s | 188 ms | 192 ms | 144 ms | – | – | 2 ms | 510 ms |
| Find one job by a unique field, p99 (lower is better) | 80 ms | 30 s | 206 ms | 205 ms | 164 ms | – | – | 290 ms | 588 ms |
| Find one job by a unique field: API (higher is better) | 2 mode | 1 mode | 1 mode | 1 mode | 2 mode | 0 mode | 0 mode | 1 mode | 1 mode |
| Find one job by a unique field: correct answers (higher is better) | 1 fraction | 0 fraction | 1 fraction | 1 fraction | 1 fraction | – | – | 1 fraction | 1 fraction |
| First page for one tenant, p50 (lower is better) | 76 ms | 30 s | 15 ms | 65 ms | 20 ms | – | – | 13 ms | 533 ms |
| First page for one tenant, p99 (lower is better) | 93 ms | 30 s | 28 ms | 70 ms | 60 ms | – | – | 29 ms | 665 ms |
| First page for one tenant: API (higher is better) | 2 mode | 1 mode | 1 mode | 1 mode | 2 mode | 0 mode | 0 mode | 1 mode | 1 mode |
| First page for one tenant: correct answers (higher is better) | 1 fraction | 0 fraction | 1 fraction | 1 fraction | 1 fraction | – | – | 1 fraction | 1 fraction |
| First page by kind and state, p50 (lower is better) | 107 ms | 1.41 s | 3 ms | 2 ms | 0.92 ms | – | 405 ms | 3 ms | 721 ms |
| First page by kind and state, p99 (lower is better) | 126 ms | 2.4 s | 4 ms | 3 ms | 2 ms | – | 433 ms | 4 ms | 903 ms |
| First page by kind and state: API (higher is better) | 2 mode | 1 mode | 1 mode | 1 mode | 2 mode | 0 mode | 1 mode | 1 mode | 1 mode |
| First page by kind and state: correct answers (higher is better) | 1 fraction | 1 fraction | 1 fraction | 1 fraction | 1 fraction | – | 1 fraction | 1 fraction | 1 fraction |
| Count by state, p50 (lower is better) | 51 ms | 56 ms | 120 ms | 1 ms | 38 ms | 0.53 ms | 431 ms | 125 ms | 1.04 s |
| Count by state, p99 (lower is better) | 58 ms | 149 ms | 147 ms | 1 ms | 48 ms | 0.94 ms | 655 ms | 138 ms | 1.19 s |
| Count by state: API (higher is better) | 2 mode | 1 mode | 1 mode | 1 mode | 1 mode | 1 mode | 1 mode | 2 mode | 1 mode |
| Count by state: correct answers (higher is better) | 1 fraction | 1 fraction | 1 fraction | 1 fraction | 1 fraction | 1 fraction | 1 fraction | 1 fraction | 1 fraction |
| Claim p99, ordinary load (lower is better) | 63 ms | 45 ms | 14 ms | 474 ms | 46 ms | 107 ms | 1.51 min | 112 ms | 29 ms |
| Claim p99 while probing (lower is better) | 17 ms | 13 ms | 13 ms | 236 ms | 43 ms | 107 ms | 3.34 min | 52 ms | 19 ms |
| Jobs/s while probing (higher is better) | 204 jobs/s | 201 jobs/s | 199 jobs/s | 200 jobs/s | 204 jobs/s | 203 jobs/s | 9.98 jobs/s | 202 jobs/s | 201 jobs/s |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗Awa 0.7 (unreleased) ↗
When an operator pauses a busy queue, does work stop everywhere, and how fast does it come back?
Setup. Two worker processes at 200 jobs/s each. The bench queue is paused; while paused, one process is restarted; then the queue is resumed. Systems without a pause stop each process's workers locally (labelled emulated), which a restarted process doesn't know about.
| System | Jobs started while paused |
|---|---|
| Awa | 0 jobs |
| PgQue | 0 jobs |
| River | 0 jobs |
| Awa 0.7 (unreleased) | 0 jobs |
| pg-boss | 7,680 jobs (emulated: each process stops) |
| Absurd | 11.6K jobs (emulated: each process stops) |
| Procrastinate | 16.2K jobs (emulated: each process stops) |
| PGMQ | 21K jobs (emulated: each process stops) |
| Oban | 21.4K jobs |
Log scale.
| System | Last job start after the pause returned |
|---|---|
| Awa | 0 ms |
| PgQue | 0 ms |
| River | 0 ms |
| Awa 0.7 (unreleased) | 0 ms |
| pg-boss | 54.9 s (emulated: each process stops) |
| Procrastinate | 55 s (emulated: each process stops) |
| Absurd | 55.5 s (emulated: each process stops) |
| PGMQ | 55.5 s (emulated: each process stops) |
| Oban | 58 s |
Log scale.
| System | First job start after the resume |
|---|---|
| pg-boss | 3 ms (emulated: each process stops) |
| Procrastinate | 8 ms (emulated: each process stops) |
| Oban | 16 ms |
| Absurd | 23 ms (emulated: each process stops) |
| Awa | 28 ms |
| River | 28 ms |
| Awa 0.7 (unreleased) | 37 ms |
| PGMQ | 53 ms (emulated: each process stops) |
| PgQue | 153 ms |
| System | Back to 90% of the earlier rate |
|---|---|
| Procrastinate | 66 ms (emulated: each process stops) |
| Oban | 174 ms |
| PGMQ | 220 ms (emulated: each process stops) |
| Absurd | 393 ms (emulated: each process stops) |
| Awa | 611 ms |
| pg-boss | 710 ms (emulated: each process stops) |
| River | 777 ms |
| PgQue | 1.02 s |
| Awa 0.7 (unreleased) | 1.31 s |
| Absurd | No pause; emulated by stopping each process's worker. |
|---|---|
| Awa | pause_queue / resume_queue, stored in the database and checked on every claim. |
| Oban | pause_queue / resume_queue, broadcast to running nodes and held in memory (open source); a restarted node starts unpaused. |
| pg-boss | No pause; emulated with offWork / work in each process. |
| PGMQ | No pause; emulated by stopping each process's readers. |
| PgQue | ticker_paused queue setting, stored in the database; batches already ticked are still delivered. |
| Procrastinate | No pause; emulated by stopping and restarting each process's worker. |
| River | QueuePause / QueueResume, stored in the database and signalled by NOTIFY. |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River | Awa 0.7 (unreleased) |
|---|---|---|---|---|---|---|---|---|---|
| Pause: API (higher is better) | 3 mode | 1 mode | 1 mode | 3 mode | 3 mode | 1 mode | 3 mode | 1 mode | 1 mode |
| Pause call (lower is better) | 0 ms | 2 ms | 0.42 ms | 0 ms | 0 ms | 4 ms | 132 ms | 9 ms | 2 ms |
| Last job start after the pause (lower is better) | 55.5 s | 0 ms | 58 s | 54.9 s | 55.5 s | 0 ms | 55 s | 0 ms | 0 ms |
| Jobs started while paused (incl. a restarted process) (lower is better) | 11.6K jobs | 0 jobs | 21.4K jobs | 7,680 jobs | 21K jobs | 0 jobs | 16.2K jobs | 0 jobs | 0 jobs |
| Resume call (lower is better) | 0 ms | 1 ms | 0.36 ms | 0 ms | 0 ms | 4 ms | 0.0048 ms | 1 ms | 2 ms |
| First job start after the resume (lower is better) | 23 ms | 28 ms | 16 ms | 3 ms | 53 ms | 153 ms | 8 ms | 28 ms | 37 ms |
| Back to 90% of the earlier rate (lower is better) | 393 ms | 611 ms | 174 ms | 710 ms | 220 ms | 1.02 s | 66 ms | 777 ms | 1.31 s |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗River ↗Awa 0.7 (unreleased) ↗
When an entire fleet dies holding about a million claimed jobs, how long until those jobs have run?
Setup. Four processes preload 1,000,000 jobs and claim them all while their handlers hold instead of running. All four are killed at once (SIGKILL); a fresh fleet of 4 × 128 workers starts, with 80 new jobs/s arriving. Every claim timeout is 300 s.
Absurdall done after 28.5 min
Awaall done after 20.5 min
Oban20% done at 36 min
pg-bossall done after 14 min
PGMQall done after 5.88 min
PgQueall done after 16.3 s
Procrastinateall done after 32 min
| System | Summary |
|---|---|
| Absurd | all done after 28.5 min |
| Awa | all done after 20.5 min |
| Oban | 20% done at 36 min |
| pg-boss | all done after 14 min |
| PGMQ | all done after 5.88 min |
| PgQue | all done after 16.3 s |
| Procrastinate | all done after 32 min |
| System | Time until every abandoned job had completed |
|---|---|
| PgQue | 16.3 s |
| PGMQ | 5.88 min |
| pg-boss | 14 min |
| Awa | 20.5 min |
| Absurd | 28.5 min |
| Procrastinate | 32 min |
| Oban | – (not all done in 30 min) |
| River | – (held no claims at the kill) |
| System | Time until the database released the dead fleet's claims |
|---|---|
| PgQue | 26 s |
| PGMQ | 3.63 min |
| pg-boss | 14.3 min |
| Awa | 22.2 min |
| Absurd | 25.6 min |
| Procrastinate | 32.1 min |
| Oban | – |
| River | – (held no claims at the kill) |
| Measurement | Absurd | Awa | Oban | pg-boss | PGMQ | PgQue | Procrastinate | River |
|---|---|---|---|---|---|---|---|---|
| Claims held at the kill (higher is better) | 998K jobs | 1.00e+3K jobs | 998K jobs | 1.1M jobs | 999K jobs | 1.00e+3K jobs | 911K jobs | – |
| First 1% done again (lower is better) | 2.19 min | 2.34 min | 2.12 min | 5.63 s | 2.05 min | 11.3 s | 5.27 min | – |
| 50% done again (lower is better) | 11.9 min | 12.3 min | – | 6.43 min | 3.71 min | 11.3 s | 19 min | – |
| All done again (lower is better) | 28.5 min | 20.5 min | – | 14 min | 5.88 min | 16.3 s | 32 min | – |
| Database claims cleared (lower is better) | 25.6 min | 22.2 min | – | 14.3 min | 3.63 min | 26 s | 32.1 min | – |
| Rescue rate, 10–90% (higher is better) | 593 jobs/s | 919 jobs/s | – | 1,533 jobs/s | 4,436 jobs/s | – | 558 jobs/s | – |
| Never completed (lower is better) | 0 jobs | 0 jobs | 798K jobs | 0 jobs | 0 jobs | 0 jobs | 0 jobs | – |
| Completed twice (lower is better) | 43 jobs | 0 jobs | 0 jobs | 80 jobs | 102 jobs | 12 jobs | 0 jobs | – |
| Fresh jobs, claim p99 during rescue (lower is better) | 1.65 min | 117 ms | 2.47 s | 43 ms | 36 ms | 106 ms | 32 ms | – |
| Postgres CPU during rescue (lower is better) | 3.39 cores | 3.89 cores | 1.87 cores | 2.88 cores | 1.3 cores | 0.16 cores | 1.89 cores | – |
Raw summaries: Absurd ↗Awa ↗Oban ↗pg-boss ↗PGMQ ↗PgQue ↗Procrastinate ↗
How the runs work
A common database does not create a common benchmark. Each system is driven through its public API by an adapter in this repository, with the same harness, workload and database.
Choose handler work, payload, offered load, pool budget and concurrency. Warm up before sampling; measure completion rate and end-to-end latency together with backlog.
Overload the producer, then reduce the load. Observe queue growth, recovery and whether retries consume capacity intended for useful work.
Hold an old snapshot, introduce long transactions and watch vacuum, WAL and lock waits. A short clean-throughput cell misses those costs.
Kill a worker, interrupt a database connection, and restart PostgreSQL in a controlled environment. Account for work and external effects, rather than only observing a returning rate.
Presence is coverage, not an endorsement or a guarantee that the adapter pin equals the current documentation snapshot.
Durable workflow
Adapter descriptor ↗Job framework
Adapter descriptor ↗Job framework
Adapter descriptor ↗Job framework
Adapter descriptor ↗Visibility queue
Adapter descriptor ↗Event queue
Adapter descriptor ↗Job framework
Adapter descriptor ↗Job framework
Adapter descriptor ↗PotatoQ, Graphile Worker, GoodJob, Que, Solid Queue and original PgQ are not presented as measured adapters. Adding them requires a version-pinned adapter and a comparable, fully specified run.
Record the database/image, host allocation, system and SDK commits, batch sizes, connections, workload and test duration. Do not combine different environments or per-system best cases into one ranking.
A per-job completion, visibility-timeout acknowledgement and whole-batch event acknowledgement have different costs. Workflow checkpointing adds another boundary. Test the mechanism you will deploy.
Missing samples stay missing. Repeat runs, preserve failures, distinguish metric sampling from independent trials, and audit reconnect/shutdown behavior before attributing an adapter result to the entire product.