Daemon benchmarks
Performance results for the Calagopus daemon, measured against Pterodactyl Wings and Pelican Wings on identical hardware. The daemon is the node agent that runs game servers: it serves the file manager and SFTP, streams archives, and reports system state. These numbers measure the daemon's own work, not how the game servers it manages perform; the daemon affects those servers only through the CPU and memory it takes from the same node.
For the web panel and its API, see panel benchmarks.
Archive and file work comes first, since those are the operations users wait on. Dots match the bar colours. How this was measured
Archive performance
↓ Lower is better - median wall clock for one run.
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 310.9 | 852.9 | 316.3 |
| EPYC 7443P | 1041.8 | 2926.2 | 1044.9 |
| 2× E5-2680v2 | 1120.4 | 4740.9 | 2426.8 |
| i5-12500 | 328.2 | 1180.4 | 454.9 |
| Altra Q80-30 | 771.3 | 2483.2 | 819.9 |
View this chart's data as a table
| System | Calagopus Wings (MiB/s) | Pelican Wings (MiB/s) | Pterodactyl Wings (MiB/s) |
|---|---|---|---|
| 9900X | 3293.9 | 1200.6 | 3237.7 |
| EPYC 7443P | 982.9 | 350.0 | 980.0 |
| 2× E5-2680v2 | 914.0 | 216.0 | 422.0 |
| i5-12500 | 3120.3 | 867.5 | 2251.4 |
| Altra Q80-30 | 1327.7 | 412.4 | 1249.0 |
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 209.6 | 367.1 | 239.8 |
| EPYC 7443P | 823.9 | 1314.1 | 996.1 |
| 2× E5-2680v2 | 1103.8 | 1993.5 | 2178.9 |
| i5-12500 | 257.9 | 477.7 | 330.4 |
| Altra Q80-30 | 719.1 | 1101.3 | 695.6 |
View this chart's data as a table
| System | Calagopus Wings (MiB/s) | Pelican Wings (MiB/s) | Pterodactyl Wings (MiB/s) |
|---|---|---|---|
| 9900X | 58.2 | 33.3 | 50.9 |
| EPYC 7443P | 14.8 | 9.3 | 12.3 |
| 2× E5-2680v2 | 11.1 | 6.1 | 5.6 |
| i5-12500 | 47.3 | 25.6 | 37.0 |
| Altra Q80-30 | 17.0 | 11.1 | 17.5 |
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 266.6 | 957.6 | 1139.4 |
| EPYC 7443P | 986.7 | 4073.0 | 5335.3 |
| 2× E5-2680v2 | 1358.7 | 7871.8 | 10189.9 |
| i5-12500 | 324.8 | 1350.5 | 1675.3 |
| Altra Q80-30 | 796.0 | 2813.1 | 3454.7 |
View this chart's data as a table
| System | Calagopus Wings (MiB/s) | Pelican Wings (MiB/s) | Pterodactyl Wings (MiB/s) |
|---|---|---|---|
| 9900X | 3841.6 | 1069.4 | 898.8 |
| EPYC 7443P | 1037.9 | 251.4 | 191.9 |
| 2× E5-2680v2 | 753.7 | 130.1 | 100.5 |
| i5-12500 | 3152.6 | 758.3 | 611.3 |
| Altra Q80-30 | 1286.5 | 364.0 | 296.4 |
Where the archive difference comes from
Mostly the deflate implementation: Calagopus uses zlib-ng and splits the work across a small thread pool (api.file_compression_threads), where both Wings builds run pgzip on one thread.
File operations
Every task scenario, timed one at a time. Median wall clock, with transfer rate, CPU cost per run and peak memory underneath.
Compression
| Scenario | Calagopus Wings | Pelican Wings | Pterodactyl Wings |
|---|---|---|---|
compress 100 MiB tree (tar.gz) | 35 ms2852 MiB/s84 cpu-ms69 MiB | 90 ms1105 MiB/s94 cpu-ms25 MiB | 61 ms1648 MiB/s116 cpu-ms54 MiB |
compress 1 GiB tree (tar.gz) | 311 ms3294 MiB/s814 cpu-ms61 MiB | 853 ms1201 MiB/s894 cpu-ms51 MiB | 316 ms3238 MiB/s762 cpu-ms66 MiB |
compress 50k small files tree (tar.gz) | 210 ms58 MiB/s251 cpu-ms47 MiB | 367 ms33 MiB/s399 cpu-ms84 MiB | 240 ms51 MiB/s315 cpu-ms36 MiB |
compress 100 MiB tree (zip) | 38 ms2620 MiB/s80 cpu-ms95 MiB | not supported | not supported |
Decompression
| Scenario | Calagopus Wings | Pelican Wings | Pterodactyl Wings |
|---|---|---|---|
decompress 100 MiB archive (tar.gz) | 45 ms2242 MiB/s89 cpu-ms68 MiB | 96 ms1042 MiB/s113 cpu-ms20 MiB | 115 ms866 MiB/s127 cpu-ms19 MiB |
decompress 1 GiB archive (tar.gz) | 433 ms2365 MiB/s931 cpu-ms52 MiB | 932 ms1099 MiB/s1070 cpu-ms20 MiB | 1,163 ms880 MiB/s1319 cpu-ms19 MiB |
decompress 50k small files archive (tar.gz) | 468 ms26 MiB/s995 cpu-ms34 MiB | 956 ms13 MiB/s1129 cpu-ms21 MiB | 1,436 ms9 MiB/s1698 cpu-ms20 MiB |
decompress 100 MiB archive (zip) | 27 ms3680 MiB/s84 cpu-ms40 MiB | 94 ms1066 MiB/s108 cpu-ms22 MiB | 118 ms848 MiB/s132 cpu-ms20 MiB |
decompress 1 GiB archive (zip) | 267 ms3842 MiB/s921 cpu-ms55 MiB | 958 ms1069 MiB/s1076 cpu-ms33 MiB | 1,139 ms899 MiB/s1284 cpu-ms32 MiB |
Deletion
| Scenario | Calagopus Wings | Pelican Wings | Pterodactyl Wings |
|---|---|---|---|
delete 100 MiB tree | 8.0 ms12 cpu-ms40 MiB | 16 ms16 cpu-ms21 MiB | 12 ms10 cpu-ms32 MiB |
delete 1 GiB tree | 59 ms114 cpu-ms32 MiB | 162 ms172 cpu-ms20 MiB | 111 ms116 cpu-ms19 MiB |
delete 50k small files tree | 85 ms172 cpu-ms31 MiB | 263 ms298 cpu-ms20 MiB | 159 ms171 cpu-ms19 MiB |
Transfer
| Scenario | Calagopus Wings | Pelican Wings | Pterodactyl Wings |
|---|---|---|---|
upload file (256 MiB) | 274 ms935 MiB/s123 cpu-ms33 MiB | 260 ms983 MiB/s89 cpu-ms20 MiB | 248 ms1033 MiB/s85 cpu-ms19 MiB |
Large directory listings
| Scenario | Calagopus Wings | Pelican Wings | Pterodactyl Wings |
|---|---|---|---|
list directory (5000 entries) | 9.6 ms17 cpu-ms56 MiB | 35 ms49 cpu-ms25 MiB | 38 ms50 cpu-ms26 MiB |
list directory (50000 entries) | 90 ms180 cpu-ms69 MiB | 318 ms389 cpu-ms86 MiB | 357 ms449 cpu-ms80 MiB |
SFTP
| Scenario | Calagopus Wings | Pelican Wings | Pterodactyl Wings |
|---|---|---|---|
sftp login | 93 ms1.350 cpu-ms31 MiB | 123 ms1.246 cpu-ms20 MiB | 123 ms1.170 cpu-ms19 MiB |
sftp list directory (5000 entries) | 113 ms22 cpu-ms31 MiB | 57 ms17 cpu-ms22 MiB | 54 ms17 cpu-ms21 MiB |
sftp list directory (50000 entries) | 450 ms235 cpu-ms30 MiB | 186 ms148 cpu-ms45 MiB | 181 ms135 cpu-ms43 MiB |
sftp upload file (256 MiB) | 409 ms625 MiB/s616 cpu-ms42 MiB | 209 ms1225 MiB/s598 cpu-ms26 MiB | 221 ms1160 MiB/s636 cpu-ms26 MiB |
sftp download file (256 MiB) | 167 ms1533 MiB/s233 cpu-ms36 MiB | 175 ms1460 MiB/s406 cpu-ms23 MiB | 185 ms1387 MiB/s419 cpu-ms22 MiB |
sftp download 100 MiB tree | 562 ms178 MiB/s334 cpu-ms33 MiB | 414 ms241 MiB/s445 cpu-ms20 MiB | 434 ms230 MiB/s456 cpu-ms19 MiB |
sftp download 50k small files tree | 7,940 ms2 MiB/s4041 cpu-ms28 MiB | 8,024 ms2 MiB/s6657 cpu-ms20 MiB | 7,685 ms2 MiB/s6375 cpu-ms20 MiB |
sftp upload 100 MiB tree | 396 ms252 MiB/s299 cpu-ms41 MiB | 608 ms164 MiB/s704 cpu-ms31 MiB | 589 ms170 MiB/s692 cpu-ms37 MiB |
sftp upload 50k small files tree | 8,078 ms2 MiB/s5396 cpu-ms30 MiB | 12,284 ms1 MiB/s10709 cpu-ms24 MiB | 12,425 ms1 MiB/s11196 cpu-ms31 MiB |
Memory
↓ Lower is better - footprint with no traffic, sampled before the run.
View this chart's data as a table
| System | Calagopus Wings (MiB) | Pelican Wings (MiB) | Pterodactyl Wings (MiB) |
|---|---|---|---|
| 9900X | 10.1 | 7.9 | 7.8 |
| EPYC 7443P | 10.1 | 7.6 | 7.4 |
| 2× E5-2680v2 | 10.0 | 7.7 | 7.9 |
| i5-12500 | 10.1 | 7.6 | 7.5 |
| Altra Q80-30 | 11.7 | 7.6 | 8.2 |
↓ Lower is better - peak footprint while serving this scenario.
View this chart's data as a table
| System | Calagopus Wings (MiB) | Pelican Wings (MiB) | Pterodactyl Wings (MiB) |
|---|---|---|---|
| 9900X | 21.6 | 79.9 | 81.0 |
| EPYC 7443P | 21.4 | 79.5 | 72.0 |
| 2× E5-2680v2 | 20.2 | 76.8 | 77.7 |
| i5-12500 | 22.1 | 80.3 | 79.7 |
| Altra Q80-30 | 20.1 | 74.4 | 77.1 |
View this chart's data as a table
| System | Calagopus Wings (MiB) | Pelican Wings (MiB) | Pterodactyl Wings (MiB) |
|---|---|---|---|
| 9900X | 50.0 | 57.3 | 59.9 |
| EPYC 7443P | 39.0 | 55.6 | 56.0 |
| 2× E5-2680v2 | 33.9 | 48.3 | 49.3 |
| i5-12500 | 44.5 | 53.8 | 54.4 |
| Altra Q80-30 | 44.6 | 57.7 | 61.2 |
View this chart's data as a table
| System | Calagopus Wings (MiB) | Pelican Wings (MiB) | Pterodactyl Wings (MiB) |
|---|---|---|---|
| 9900X | 62.7 | 39.7 | 41.1 |
| EPYC 7443P | 45.9 | 40.6 | 37.3 |
| 2× E5-2680v2 | 38.5 | 40.0 | 43.4 |
| i5-12500 | 59.7 | 39.8 | 39.9 |
| Altra Q80-30 | 61.0 | 41.2 | 41.3 |
View this chart's data as a table
| System | Calagopus Wings (MiB) | Pelican Wings (MiB) | Pterodactyl Wings (MiB) |
|---|---|---|---|
| 9900X | 181.8 | 42.4 | 44.0 |
| EPYC 7443P | 166.1 | 43.2 | 41.4 |
| 2× E5-2680v2 | 171.0 | 43.1 | 44.4 |
| i5-12500 | 171.5 | 44.3 | 44.3 |
| Altra Q80-30 | 216.1 | 46.0 | 46.0 |
View this chart's data as a table
| System | Calagopus Wings (MiB) | Pelican Wings (MiB) | Pterodactyl Wings (MiB) |
|---|---|---|---|
| 9900X | 53.0 | 43.3 | 35.4 |
| EPYC 7443P | 46.6 | 39.3 | 37.0 |
| 2× E5-2680v2 | 46.1 | 39.4 | 38.1 |
| i5-12500 | 53.2 | 41.5 | 35.0 |
| Altra Q80-30 | 76.4 | 40.1 | 37.4 |
Request scenarios
Why this matters on shared hosting
The daemon competes with your game servers for the same CPU. A game server that misses its CPU time drops ticks instead of slowing down gradually, so how cheaply the daemon handles each operation matters more than its peak throughput.
Throughput
↑ Higher is better - more requests served per second.
View this chart's data as a table
| System | Calagopus Wings (req/s) | Pelican Wings (req/s) | Pterodactyl Wings (req/s) |
|---|---|---|---|
| 9900X | 354577.0 | 3745.1 | 4260.1 |
| EPYC 7443P | 131777.7 | 2139.4 | 2180.3 |
| 2× E5-2680v2 | 80985.5 | 1397.6 | 1407.8 |
| i5-12500 | 213747.2 | 2487.4 | 2526.6 |
| Altra Q80-30 | 106901.8 | 1792.0 | 1831.0 |
View this chart's data as a table
| System | Calagopus Wings (req/s) | Pelican Wings (req/s) | Pterodactyl Wings (req/s) |
|---|---|---|---|
| 9900X | 34075.8 | 11866.2 | 10614.4 |
| EPYC 7443P | 10588.9 | 4320.1 | 3967.3 |
| 2× E5-2680v2 | 4659.1 | 2830.1 | 2511.4 |
| i5-12500 | 25061.7 | 8126.6 | 7339.3 |
| Altra Q80-30 | 14593.9 | 4324.0 | 3964.2 |
View this chart's data as a table
| System | Calagopus Wings (req/s) | Pelican Wings (req/s) | Pterodactyl Wings (req/s) |
|---|---|---|---|
| 9900X | 109605.0 | 119205.9 | 120649.9 |
| EPYC 7443P | 46487.9 | 41148.2 | 40185.7 |
| 2× E5-2680v2 | 21606.3 | 25010.0 | 26038.1 |
| i5-12500 | 74996.7 | 76616.7 | 78662.0 |
| Altra Q80-30 | 37923.7 | 32533.8 | 34202.5 |
View this chart's data as a table
| System | Calagopus Wings (req/s) | Pelican Wings (req/s) | Pterodactyl Wings (req/s) |
|---|---|---|---|
| 9900X | 768.3 | 1068.1 | 1059.7 |
| EPYC 7443P | 546.7 | 376.8 | 374.8 |
| 2× E5-2680v2 | 461.3 | 196.3 | 242.2 |
| i5-12500 | 524.5 | 463.4 | 475.7 |
| Altra Q80-30 | 699.3 | 356.9 | 365.8 |
View this chart's data as a table
| System | Calagopus Wings (req/s) | Pelican Wings (req/s) | Pterodactyl Wings (req/s) |
|---|---|---|---|
| 9900X | 135541.1 | 160807.5 | 189511.6 |
| EPYC 7443P | 65999.5 | 61265.7 | 69935.5 |
| 2× E5-2680v2 | 33833.9 | 35013.3 | 41084.1 |
| i5-12500 | 79237.4 | 86753.5 | 104315.0 |
| Altra Q80-30 | 48251.6 | 49177.2 | 54234.1 |
Average latency
↓ Lower is better - faster average response.
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 1.1 | 101.8 | 89.7 |
| EPYC 7443P | 2.9 | 178.5 | 174.8 |
| 2× E5-2680v2 | 4.7 | 272.1 | 270.8 |
| i5-12500 | 1.8 | 153.7 | 151.3 |
| Altra Q80-30 | 3.6 | 212.6 | 208.5 |
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 11.2 | 32.2 | 36.0 |
| EPYC 7443P | 36.1 | 88.2 | 96.0 |
| 2× E5-2680v2 | 81.8 | 134.3 | 151.2 |
| i5-12500 | 15.3 | 47.0 | 52.1 |
| Altra Q80-30 | 26.2 | 88.0 | 96.1 |
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 3.5 | 3.2 | 3.2 |
| EPYC 7443P | 8.2 | 9.3 | 9.5 |
| 2× E5-2680v2 | 17.7 | 15.3 | 14.7 |
| i5-12500 | 5.1 | 5.0 | 4.9 |
| Altra Q80-30 | 10.1 | 11.7 | 11.2 |
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 493.3 | 355.3 | 356.5 |
| EPYC 7443P | 690.8 | 989.4 | 995.5 |
| 2× E5-2680v2 | 814.4 | 1900.2 | 1537.5 |
| i5-12500 | 720.8 | 810.6 | 789.7 |
| Altra Q80-30 | 540.1 | 1045.5 | 1035.4 |
View this chart's data as a table
| System | Calagopus Wings (ms) | Pelican Wings (ms) | Pterodactyl Wings (ms) |
|---|---|---|---|
| 9900X | 2.8 | 2.4 | 2.0 |
| EPYC 7443P | 5.8 | 6.2 | 5.5 |
| 2× E5-2680v2 | 11.3 | 10.9 | 9.3 |
| i5-12500 | 4.8 | 4.4 | 3.7 |
| Altra Q80-30 | 7.9 | 7.8 | 7.0 |
CPU cost per request
↓ Lower is better - less CPU burned per response, with the idle floor removed.
View this chart's data as a table
| System | Calagopus Wings (cpu-ms) | Pelican Wings (cpu-ms) | Pterodactyl Wings (cpu-ms) |
|---|---|---|---|
| 9900X | 0.0180 | 0.462 | 0.420 |
| EPYC 7443P | 0.0450 | 0.827 | 0.778 |
| 2× E5-2680v2 | 0.0830 | 1.5 | 1.5 |
| i5-12500 | 0.0200 | 0.588 | 0.543 |
| Altra Q80-30 | 0.0550 | 0.973 | 0.920 |
View this chart's data as a table
| System | Calagopus Wings (cpu-ms) | Pelican Wings (cpu-ms) | Pterodactyl Wings (cpu-ms) |
|---|---|---|---|
| 9900X | 0.234 | 0.638 | 0.714 |
| EPYC 7443P | 0.753 | 1.8 | 1.9 |
| 2× E5-2680v2 | 1.7 | 2.7 | 3.1 |
| i5-12500 | 0.317 | 0.956 | 1.1 |
| Altra Q80-30 | 0.546 | 1.6 | 1.8 |
View this chart's data as a table
| System | Calagopus Wings (cpu-ms) | Pelican Wings (cpu-ms) | Pterodactyl Wings (cpu-ms) |
|---|---|---|---|
| 9900X | 0.0670 | 0.0570 | 0.0560 |
| EPYC 7443P | 0.171 | 0.169 | 0.174 |
| 2× E5-2680v2 | 0.368 | 0.289 | 0.285 |
| i5-12500 | 0.0890 | 0.0860 | 0.0840 |
| Altra Q80-30 | 0.196 | 0.201 | 0.196 |
View this chart's data as a table
| System | Calagopus Wings (cpu-ms) | Pelican Wings (cpu-ms) | Pterodactyl Wings (cpu-ms) |
|---|---|---|---|
| 9900X | 8.0 | 6.7 | 6.6 |
| EPYC 7443P | 10.2 | 18.1 | 18.3 |
| 2× E5-2680v2 | 11.1 | 34.3 | 28.8 |
| i5-12500 | 8.9 | 12.1 | 12.0 |
| Altra Q80-30 | 7.5 | 18.2 | 17.7 |
View this chart's data as a table
| System | Calagopus Wings (cpu-ms) | Pelican Wings (cpu-ms) | Pterodactyl Wings (cpu-ms) |
|---|---|---|---|
| 9900X | 0.0540 | 0.0430 | 0.0360 |
| EPYC 7443P | 0.120 | 0.113 | 0.0990 |
| 2× E5-2680v2 | 0.234 | 0.206 | 0.179 |
| i5-12500 | 0.0770 | 0.0720 | 0.0600 |
| Altra Q80-30 | 0.152 | 0.136 | 0.125 |
Test setup
Methodology
All three daemons run from their official :latest Docker images with no configuration beyond initial setup, driven by our open-source benchmarking suite at CPU limits of 1, 2, 4 and 8 CPUs.
Request scenarios hold 384 concurrent connections for 10 seconds after a 1-second warmup and report throughput and latency from a single run. The suite defaults to 32 connections (8 for the 8 MiB download) and this run passed -c 384, so its README figures differ from these. Task scenarios are timed one at a time, 5 times (10 for directory listings and SFTP logins) after a warmup run, and report the median. The gap between a cell's fastest and slowest run is typically 9% of its median and above 17% for a quarter of the cells, so treat task gaps under about 20% as ties.
Files live on a tmpfs. Each daemon's data and temp directories are RAM-backed, and fixtures are copied in before each task outside the timed window, so writes never reach a disk and fsync costs nothing. File tasks therefore measure the daemon rather than the drive: archive and listing tasks compare codecs and serialisation, not disk reads. On an HDD, a network volume or a cold cache, storage would dominate and the gaps would shrink.
SFTP scenarios use the OpenSSH sftp client in batch mode with aes128-gcm@openssh.com and an ed25519 key. Transfers and listings share one multiplexed connection, so they exclude the handshake; sftp login opens a fresh connection each run, which includes the daemon's key check against the suite's mock panel API.
Memory is anonymous memory (anon), excluding page cache and slab. Shared memory is left out because on a tmpfs it is just the file pages the daemon wrote. cpu-ms is CPU time per request or task run with the idle floor subtracted, and lower is better.
Daemons tested
1.2.81.0.0-rc.11.13.4Capability gaps
Where a daemon cannot perform a scenario the suite skips it rather than recording a zero, so these are missing features, not missing measurements.
compress 100 MiB tree (zip) is not supported by Pelican Wings, Pterodactyl Wings
Data
Where the numbers come from
Every figure here comes from the suite's JSON reports (pnpm run bench <target> --json), checked into the site repository as one typed file per daemon and system, with each scenario's load profile, sampled CPU and memory, and the suite's notes. The import script rounds to three decimals and drops per-run timings and failure strings. The reports record each daemon's self-reported version, but not the suite commit or image digests.