windcloud / enge-sande — VPS-L

€19/mo · storage net-slow · host 3637eb36eb5f · 2026-07-17T15:22:48Z · submitter uptimeify

Profiles

profile	grade	bound by
postgres_oltp	F	`disk.wal_fsync.p999_us`
timescale_ingest	F	`disk.wal_fsync.p999_us`
patroni_member	F	`disk.wal_fsync.p999_us`
redis_sentinel	F	`cpu.single_thread_eps`
worker_probe	F	`cpu.single_thread_eps`
playwright_node	F	`cpu.single_thread_eps`
nuxt_ssr	F	`cpu.single_thread_eps`

Tail-latency reality

fsync cumulative distribution

hover the curve — snaps to measured percentiles

Points are the measured p50 / p99 / p99.9 / max — the line between them is interpolated (we store percentiles, not a full histogram). Fixed axis: a tight curve on the left is a good host; a curve swept right is an exploding tail.

marketing claim vs measured

− typical marketing claim + what p99bench measured here

−"Up to N,000 IOPS"

+ fsync p99.9 latency at QD1 242 ms F

−"NVMe SSD — low latency"

+ p99.9 vs median, not the mean p50 0.77 ms → p99.9 21.4 ms C

−"Dedicated vCPUs"

+ CPU steal measured under full load 0.07% steal

−"High-performance cores"

+ single-thread throughput 399 eps F

−"Fast" (short benchmark)

+ IOPS drop over 30 min sustained load 9.04% degradation

−"No noisy neighbours"

+ scheduler stall p99.9 0.24 ms B

Disk

bound by disk.wal_fsync.p999_us

WAL fsync latency ladder (ms, QD1) Every COMMIT waits on one fdatasync. p99.9 is the graded metric — the tail that stalls a database.

avg

9.3 ms

p50

7.9 ms

p99

23.7 ms

p99.9

242 ms F

max

372 ms

8K random latency p99.9 by operation (ms)

read QD1

21.4 ms

read

16.9 ms

write

249 ms

mixed

32.6 ms

Throughput

seq read

3,436 MB/s

seq write

584 MB/s

Random IOPS

read 8K

66,259 IOPS

write 8K

12,832 IOPS

mixed 8K

26,142 IOPS

fsync

103 IOPS

Sustained IOPS — first vs last minute of 30 min 9.04% degradation over 30 min

first min

36,869 IOPS

last min

33,538 IOPS

exact values & per-metric grades

metric	value	grade
`disk.wal_fsync.p999_us`	242221.06	F
`disk.wal_fsync.iops`	103.35	D
`disk.rand_read_8k_qd1.p99_us`	3817.47	C
`disk.rand_read_8k.iops`	66258.92	B
`disk.rand_write_8k.iops`	12832.41	C
`disk.seq_write.bw_mbs`	583.55	B
`disk.seq_read.bw_mbs`	3435.73	A
`disk.steady_state.degradation_pct`	9.04	B

CPU

bound by cpu.single_thread_eps

Throughput (events/s)

single

399 eps F

multi

1,551 eps

Scheduler stall ladder (ms) How long the kernel keeps a runnable thread waiting. p99.9 is graded.

p99

0.1 ms

p99.9

0.24 ms B

max

4.9 ms

Sustained throughput — first vs last minute 2% degradation · steal 0.2%

first min

1,605 eps

last min

1,565 eps

exact values & per-metric grades

metric	value	grade
`cpu.single_thread_eps`	399.16	F
`cpu.scaling_efficiency`	0.971	A
`cpu.steal_pct_under_load`	0.07	A
`cpu.stall_p999_us`	242	B
`cpu.steady_state.degradation_pct`	2	A
`cpu.tls_verify_s`	9439.5	C

RAM

bound by ram.bw_read_mbs

Memory bandwidth (MB/s) bw read is measured outside cache (the graded metric); seq/rnd are raw STREAM-style figures.

bw read

30,402 MB/s B

seq read

63,730 MB/s

seq write

47,376 MB/s

rnd read

8,039 MB/s

rnd write

768 MB/s

exact values & per-metric grades

metric	value	grade
`ram.bw_read_mbs`	30401.58	B

Network

bound by network.loss_pct

Round-trip p99 by target (ms) Same fixed target list from every host — distance is constant, so peering and neighbour noise surface.

hetzner-fsn1

17.1 ms

hetzner-hel1

29.5 ms

hetzner-ash

93.6 ms

Packet loss by target (%)

hetzner-fsn1

0 %

hetzner-hel1

0 %

hetzner-ash

0 %

exact values & per-metric grades

metric	value	grade
`network.loss_pct`	0	A
`network.rtt_jitter_ratio`	1.0068259385665528	A