Benchmarks

Performance results for Calagopus, measured against Pterodactyl on identical hardware. Each chart compares both panels across every test configuration.

peak throughput

263k req/s

Calagopus

799 Pterodactyl

avg response

1.9 ms

Calagopus

627 ms Pterodactyl

idle memory

138 MiB

Calagopus

216 MiB Pterodactyl

Methodology

Tests use Calagopus 1.0.3 and Pterodactyl 1.12.2, both running from their official Docker images with no additional configuration beyond initial setup. Each test ran oha with 500 concurrent connections for 60 seconds, from a separate machine on the same LAN over a 10 Gbps link. Two endpoints were targeted: the panel root (/) and an authenticated API endpoint (/api/client/permissions). Both panels used default rate limiting; the high [429] counts on the API endpoint are expected.

Configurations are labeled <CPU> · <threads>t · <RAM>G. Memory figures come from Proxmox LXC container stats (the Docker container runs inside the LXC).

Test environments

Ryzen 9 9900X

CPUAMD Ryzen 9 9900X

RAMDDR5-6000

StorageRAID 1 NVMe

KernelLinux 6.17.4-2-pve

c7s idle0.05%c7s peak78–90%ptero peak56–97%

EPYC 7443P

CPUAMD EPYC 7443P

RAMDDR4-2666

StorageRAID 1 NVMe

KernelLinux 6.17.4-2-pve

c7s idle0.06%c7s peak91–99%ptero peak57–98%

2× Xeon E5-2680 v2

CPU2× Intel Xeon E5-2680 v2

RAMDDR3-1600

StorageRAID 1 SATA SSD

KernelLinux 6.17.13-2-pve

c7s idle0.12%c7s peak85–99%ptero peak60–100%

Ampere Altra Q80-30

CPUAmpere Altra Q80-30

RAMDDR4-2133

StorageRAID 1 NVMe

KernelLinux 6.12.63-rt-arm64

c7s idle0.08%c7s peak17–93%ptero peak52–99%

Memory usage

Idle

Peak under load

Calagopus idles around 140-160 MiB regardless of host. Under load it rarely exceeds 300 MiB, while Pterodactyl peaks above 1 GiB on the 9900X.

Throughput

Endpoint: /

Endpoint: /api/client/permissions

Average latency

Endpoint: /

Endpoint: /api/client/permissions

p99 latency

Endpoint: /

Endpoint: /api/client/permissions