Server benchmark

server_benchmark.md

Results

Server	Price*	CPU (1 thread)	CPU (4 threads)	IO
Scaleway Start1-XS Atom C3955, 25G NVMe	$4	21 sec P₉₅ = 2.1ms	21 sec P₉₅ = 14ms	🚗 653 IO/s, 10 Mb/sec P₉₅ = 0.40ms
Scaleway VC1-S Atom C2750, 50G SSD	$4	46 sec ⚠️ P₉₅ = 4.6ms	23 sec P₉₅ = 14ms	🚲 289 IO/s, 4.5 Mb/sec P₉₅ = 0.39ms
Hetzner CPX11 AMD EPYC, 40G SSD	$5	20 sec P₉₅ = 3.2ms	9.6 sec P₉₅ = 11ms	✈️ 3656 IO/s, 57 Mb/sec P₉₅ = 0.29ms
Hetzner CX21 Intel Xeon, 40G SSD	$6	24 sec P₉₅ = 2.8ms	12.5 sec P₉₅ = 11ms	🚤 1946 IO/s, 30 Mb/sec P₉₅ = 0.97ms
AWS t2.micro 1G Intel E5-2676 v3 2.4GHz, EBS	$6	29 sec P₉₅ = 3.1ms	29 sec P₉₅ = 15ms	🚤 1150 IO/s, 17 Mb/sec P₉₅ = 1.01ms
Digital Ocean 1G Xeon E5-2650, 25G SSD	$6	31 sec P₉₅ = 3ms	31 sec P₉₅ = 19ms	✈️ 3412 IO/s, 53 Mb/sec P₉₅ = 0.52ms
OVH VPS SSD2 Xeon E3-12xx, 20G SSD	$7	31 sec P₉₅ = 3.4ms	31 sec P₉₅ = 15ms	🚤 1113 IO/s, 17 Mb/sec P₉₅ = 0.68ms
Hetzner CPX21 AMD EPYC, 80G SSD	$8	18 sec P₉₅ = 3ms	6.9 sec P₉₅ = 9.6ms	🚤 1950 IO/s, 30 Mb/sec P₉₅ = 1.08ms
OVH VPS CLOUD RAM1 Xeon E3-12xx, 25G NVMe HA	$11	31 sec P₉₅ = 3.4ms	32 sec P₉₅ = 15ms	🚲 110 IO/s, 1.7 Mb/sec P₉₅ = 8ms
Hetzner CX31 Intel Xeon, 80G SSD	$11	30 sec P₉₅ = 3.5ms	15 sec P₉₅ = 11.45ms	🚗 900 IO/s, 14 Mb/sec P₉₅ = 2.57ms
Linode 2048 Xeon E5-2680, 24G SSD	$12	32 sec P₉₅ = 3.4ms	32 sec P₉₅ = 14ms	✈️ 3426 IO/s, 53 Mb/sec P₉₅ = 0.15ms
SSDNode 24G Intel 2.6GHz, 120G SSD	$150/y	46 sec ⚠️ P₉₅ = 5.1ms	50 sec ⚠️ P₉₅ = 86ms	🚗 660 IO/s, 10 Mb/sec P₉₅ = 2ms
Scaleway Start1-L Atom C3955, 50G NVMe	$14	21 sec P₉₅ = 2.2ms	5.4 sec P₉₅ = 2.3ms	🚗 626 IO/s, 9.8 Mb/sec P₉₅ = 0.45ms
Hetzner CPX31 AMD EPYC, 160G SSD	$15	17 sec P₉₅ = 3ms	5.3 sec P₉₅ = 3ms	✈️ 2560 IO/s, 40 Mb/sec P₉₅ = 0.74ms
Digital Ocean 1G/3vCPU Xeon E5-2650L, 60G SSD	$18	39 sec P₉₅ = 4ms	13 sec P₉₅ = 12ms	✈️ 2644 IO/s, 41 Mb/sec P₉₅ = 0.57ms
Hetzner CX41 Intel Xeon, 160G SSD	$19	26 sec P₉₅ = 2.8ms	6.9 sec P₉₅ = 2.9ms	🚤 1700 IO/s, 26 Mb/sec P₉₅ = 1.06ms
Digital Ocean 4G Xeon Gold 6140, 80G SSD	$24	25 sec P₉₅ = 2.8ms	13 sec P₉₅ = 11ms	✈️ 2900 IO/s, 45 Mb/sec P₉₅ = 0.4ms
Scaleway C2M Atom C2750, 50G SSD	$24	46 sec ⚠️ P₉₅ = 4.6ms	11 sec P₉₅ = 4.5ms	🚗 500 IO/s, 7.8 Mb/sec P₉₅ = 0.02ms
Vultr Compute 4G Intel 2.6GHz, 80G SSD	$24	25 sec P₉₅ = 2.8ms	12.8 sec P₉₅ = 11ms	✈️ 2800 IO/s, 43 Mb/sec P₉₅ = 0.58ms
Hetzner CCX11 Intel Xeon, 80G SSD	$27	28 sec P₉₅ = 4.4ms	16 sec P₉₅ = 13.5ms	🚤 2100 IO/s, 31 Mb/sec P₉₅ = 0.97ms
Hetzner CCX12 AMD EPYC, 80G SSD	$27	6.9 sec ❓ P₉₅ = 0.7ms	6.2 sec P₉₅ = 9.22ms	🚀 8770 IO/s, 107 Mb/sec P₉₅ = 0.22ms
Vultr High CPU 4G Intel 3.8GHz, 128G NVMe	$28	21 sec P₉₅ = 2.4ms	10.8 sec P₉₅ = 10ms	🚀 7500 IO/s, 117 Mb/sec P₉₅ = 0.17ms
Online Start-2-M-SSD Atom C2750, 250G SSD	$28	46 sec ⚠️ P₉₅ = 4.6ms	11 sec P₉₅ = 4.5ms	🚗 789 IO/s, 12.3 Mb/sec P₉₅ = 0.02ms
OVH Public Cloud B2-7 Intel 2.2Ghz, 50G SSD	$30	28 sec P₉₅ = 2.9ms	14 sec P₉₅ = 11ms	✈️ 3020 IO/s, 47 Mb/sec P₉₅ = 0.33ms
OVH kimsufi i5 3570s, 2T SATA HDD	$34	23 sec P₉₅ = 2.4ms	6.4 sec P₉₅ = 2.7ms	🐢 17 IO/s, 0.3 Mb/sec P₉₅ = 30ms
OVH Public Cloud C2-7 Intel 3.1Ghz, 50G SSD	$43	25 sec P₉₅ = 2.7ms	12 sec P₉₅ = 11ms	✈️ 3028 IO/s, 47 Mb/sec P₉₅ = 0.67ms
Digital Ocean 8G Xeon Gold 6140, 160G SSD	$48	25 sec P₉₅ = 2.8ms	6.7 sec P₉₅ = 3.3ms	✈️ 3800 IO/s, 59 Mb/sec P₉₅ = 0.58ms
SoYouStart E3-SSD-3 E3 1245v2, 3x 120 Go SSD	$50	22 sec P₉₅ = 2.2ms	5.9 sec P₉₅ = 2.3ms	🚀 6239 IO/s, 97 Mb/sec P₉₅ = 0.01ms
NUC8i5BEH 16G i5-8259u, 256Gnvme	$600 buy	19 sec P₉₅ = 1.9ms	5.0 sec P₉₅ = 2.0ms	✈️ 2300 IO/s, 32 Mb/sec P₉₅ = 0.1ms

*VAT included

---

Testing protocol

sysbench 1.0 (ubuntu 18.04+)

sudo apt install sysbench
# CPU
sysbench cpu --cpu-max-prime=20000 --time=0 --events=10000 run
sysbench cpu --cpu-max-prime=20000 --time=0 --events=10000 --num-threads=4 run
# IO
sudo sysbench fileio --file-total-size=10G prepare
sudo sysbench fileio --file-total-size=10G --file-test-mode=rndrw --time=60 --max-requests=0 run
sudo sysbench fileio --file-total-size=10G cleanup

Much faster IO test

sudo sysbench fileio --file-total-size=10G --file-num=1 prepare
sudo sysbench fileio --file-total-size=10G --file-test-mode=rndrw --time=60 --file-io-mode=async --file-extra-flags=direct --file-num=1 --file-rw-ratio=1 --file-fsync-freq=0 --max-requests=0 run
sudo sysbench fileio --file-total-size=10G --file-num=1 cleanup

sysbench 0.4 (ubuntu 16.04)

sudo apt install sysbench
# CPU
sysbench --test=cpu --cpu-max-prime=20000 run
sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run
# IO
sudo sysbench --test=fileio --file-total-size=10G prepare
sudo sysbench --test=fileio --file-total-size=10G --file-test-mode=rndrw --init-rng=on --max-time=60 --max-requests=0 run
sudo sysbench --test=fileio --file-total-size=10G cleanup

Here is the AWS t2.micro instance benchmark: https://gist.github.com/spuyet/b43c5258bedda772bb1280d14479914b

The pricing varies following the selected plan (dedicated vs on demand instance, 1 year vs 3 years contract, ...): https://aws.amazon.com/ec2/pricing/
FTR TVA is excluded.

Hetzner CCX12
AMD Epyc 7003, 80GB SSD
25.40 EUR/mo

CPU (1 thread)
6.9225s
P95=0.7ms

CPU (4 thread)
6.2795s
P95=9.22ms

IO
8770.96 fsyncs
64.24MiB/s read, 42.93 MiB/s write
P95 = 0.22ms

root@ubuntu-8gb-nbg1-5:~# sysbench cpu --cpu-max-prime=20000 --time=0 --events=10000 run

sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 20000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:  1444.47

General statistics:
    total time:                          6.9225s
    total number of events:              10000

Latency (ms):
         min:                                    0.68
         avg:                                    0.69
         max:                                    1.56
         95th percentile:                        0.70
         sum:                                 6921.06

Threads fairness:
    events (avg/stddev):           10000.0000/0.00
    execution time (avg/stddev):   6.9211/0.00


root@ubuntu-8gb-nbg1-5:~# sysbench cpu --cpu-max-prime=20000 --time=0 --events=10000 --num-threads=4 run
WARNING: --num-threads is deprecated, use --threads instead
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 4
Initializing random number generator from current time


Prime numbers limit: 20000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:  1592.37

General statistics:
    total time:                          6.2795s
    total number of events:              10000

Latency (ms):
         min:                                    0.90
         avg:                                    2.51
         max:                                   17.22
         95th percentile:                        9.22
         sum:                                25094.04

Threads fairness:
    events (avg/stddev):           2500.0000/3.74
    execution time (avg/stddev):   6.2735/0.00

root@ubuntu-8gb-nbg1-5:~# sudo sysbench fileio --file-total-size=10G --file-test-mode=rndrw --time=60 --max-requests=0 run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Extra file open flags: (none)
128 files, 80MiB each
10GiB total file size
Block size 16KiB
Number of IO requests: 0
Read/Write ratio for combined random IO test: 1.50
Periodic FSYNC enabled, calling fsync() each 100 requests.
Calling fsync() at the end of test, Enabled.
Using synchronous I/O mode
Doing random r/w test
Initializing worker threads...

Threads started!


File operations:
    reads/s:                      4111.29
    writes/s:                     2740.86
    fsyncs/s:                     8770.96

Throughput:
    read, MiB/s:                  64.24
    written, MiB/s:               42.83

General statistics:
    total time:                          60.0098s
    total number of events:              937421

Latency (ms):
         min:                                    0.00
         avg:                                    0.06
         max:                                    9.59
         95th percentile:                        0.22
         sum:                                59759.54

Threads fairness:
    events (avg/stddev):           937421.0000/0.00
    execution time (avg/stddev):   59.7595/0.00