Aeza AMD EPYC 8534P Storage VPS Benchmark#

Russian provider Aeza has rolled out a new storage product line today, and it’s pricey. The VPS uses an AMD EPYC 8534P CPU with a ZFS RAID storage array and time-based bandwidth throttling. Current plans start at 4c/4G/4T for USD $40 per month. At this price point, not including an NVMe disk as the system drive is really hard to justify, and the overall price‑performance ratio is quite poor.

From 14:00 to 01:00 Beijing time, the speed can reach 10 Gbps.
From 01:00 to 14:00 Beijing time, the speed is capped at 1 Gbps.

Benchmark#

CPU#

1
root@catcat:~# lscpu
2
Architecture:             x86_64
3
  CPU op-mode(s):         32-bit, 64-bit
4
  Address sizes:          48 bits physical, 57 bits virtual
5
  Byte Order:             Little Endian
6
CPU(s):                   4
7
  On-line CPU(s) list:    0-3
8
Vendor ID:                AuthenticAMD
9
  BIOS Vendor ID:         Red Hat
10
  Model name:             AMD EPYC 8534P 64-Core Processor
11
    BIOS Model name:      RHEL 7.6.0 PC (i440FX + PIIX, 1996)  CPU @ 2.0GHz
12
    BIOS CPU family:      1
13
    CPU family:           25
14
    Model:                160
15
    Thread(s) per core:   1
16
    Core(s) per socket:   4
17
    Socket(s):            1
18
    Stepping:             2
19
    BogoMIPS:             4593.09
20
    Flags:                fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflus                          h mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm rep_good n                          opl cpuid extd_apicid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1
21
                          sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hyperv                          isor lahf_lm cmp_legacy svm cr8_legacy abm sse4a misalignsse 3dnowprefetch osv                          w perfctr_core invpcid_single ssbd ibrs ibpb stibp vmmcall fsgsbase tsc_adjust                           bmi1 avx2 smep bmi2 erms invpcid avx512f avx512dq rdseed adx smap avx512ifma
22
                          clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsav                          es avx512_bf16 clzero xsaveerptr wbnoinvd arat npt lbrv nrip_save tsc_scale vm                          cb_clean flushbyasid pausefilter pfthreshold v_vmsave_vmload vgif avx512vbmi u                          mip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx5                          12_vpopcntdq la57 rdpid fsrm flush_l1d arch_capabilities
23
Virtualization features:
24
  Virtualization:         AMD-V
25
  Hypervisor vendor:      KVM
26
  Virtualization type:    full
27
Caches (sum of all):
28
  L1d:                    256 KiB (4 instances)
29
  L1i:                    256 KiB (4 instances)
30
  L2:                     2 MiB (4 instances)
31
  L3:                     64 MiB (4 instances)
32
NUMA:
33
  NUMA node(s):           1
34
  NUMA node0 CPU(s):      0-3
35
Vulnerabilities:
36
  Gather data sampling:   Not affected
37
  Itlb multihit:          Not affected
38
  L1tf:                   Not affected
39
  Mds:                    Not affected
40
  Meltdown:               Not affected
41
  Mmio stale data:        Not affected
42
  Reg file data sampling: Not affected
43
  Retbleed:               Not affected
44
  Spec rstack overflow:   Mitigation; safe RET
45
  Spec store bypass:      Mitigation; Speculative Store Bypass disabled via prctl
46
  Spectre v1:             Mitigation; usercopy/swapgs barriers and __user pointer sanitization
47
  Spectre v2:             Mitigation; Retpolines; IBPB conditional; IBRS_FW; STIBP disabled; RSB filling                          ; PBRSB-eIBRS Not affected; BHI Not affected
48
  Srbds:                  Not affected
49
  Tsx async abort:        Not affected

Yabs#

1
Basic System Information:
2
---------------------------------
3
Uptime     : 0 days, 0 hours, 0 minutes
4
Processor  : AMD EPYC 8534P 64-Core Processor
5
CPU cores  : 4 @ 2296.548 MHz
6
AES-NI     : ✔️ Enabled
7
VM-x/AMD-V : ✔️ Enabled
8
RAM        : 3.8 GiB
9
Swap       : 512.0 MiB
10
Disk       : 3.9 TiB
11
Distro     : Debian GNU/Linux 12 (bookworm)
12
Kernel     : 6.1.0-22-amd64
13
VM Type    : KVM
14
IPv4/IPv6  : ✔️ Online / ❌ Offline
15

16
IPv4 Network Information:
17
---------------------------------
18
ISP        : Aeza International LTD
19
ASN        : AS210644 AEZA INTERNATIONAL LTD
20
Host       : Aeza International LTD
21
Location   : London, England (ENG)
22
Country    : United Kingdom
23

24
fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/vda2):
25
---------------------------------
26
Block Size | 4k            (IOPS) | 64k           (IOPS)
27
  ------ | --- ---- | ---- ----
28
Read       | 3.93 MB/s      (984) | 58.09 MB/s     (907)
29
Write      | 3.96 MB/s      (990) | 58.66 MB/s     (916)
30
Total      | 7.89 MB/s     (1.9k) | 116.75 MB/s   (1.8k)
31
           |                      |
32
Block Size | 512k          (IOPS) | 1m            (IOPS)
33
  ------ | --- ---- | ---- ----
34
Read       | 410.54 MB/s    (801) | 512.77 MB/s    (500)
35
Write      | 432.36 MB/s    (844) | 546.92 MB/s    (534)
36
Total      | 842.90 MB/s   (1.6k) | 1.05 GB/s     (1.0k)
37

38
iperf3 Network Speed Tests (IPv4):
39
---------------------------------
40
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping
41
----- | ----- | ---- | ---- | ----
42
Clouvider       | London, UK (10G)          | 8.97 Gbits/sec  | 6.98 Gbits/sec  | 6.10 ms
43
Eranium         | Amsterdam, NL (100G)      | busy            | busy            | --
44
Uztelecom       | Tashkent, UZ (10G)        | 1.64 Gbits/sec  | 1.90 Gbits/sec  | 95.9 ms
45
Leaseweb        | Singapore, SG (10G)       | 805 Mbits/sec   | 987 Mbits/sec   | 178 ms
46
Clouvider       | Los Angeles, CA, US (10G) | busy            | 2.60 Mbits/sec  | --
47
Leaseweb        | NYC, NY, US (10G)         | 2.11 Gbits/sec  | 2.45 Gbits/sec  | 74.6 ms
48
Edgoo           | Sao Paulo, BR (1G)        | 880 Mbits/sec   | 1.08 Gbits/sec  | 165 ms
49

50
Geekbench 6 BenchMark Test:
51
---------------------------------
52
Test            | Value
53
                |
54
Single Core     | 1736
55
Multi Core      | 5514
56
Full Test       | https://browser.geekbench.com/v6/cpu/6834382
57

58
YABS completed in 11 min 58 sec

GeekBench5#

1
Geekbench 5 测试结果
2

3
系统信息
4
  Operating System              Debian GNU/Linux 12 (bookworm)
5
  Kernel                        Linux 6.1.0-22-amd64 x86_64
6
  Model                         Red Hat KVM
7
  Motherboard                   N/A
8
  BIOS                          SeaBIOS 1.16.0-4.module_el8.9.0+3659+9c8643f3
9

10
处理器信息
11
  Name                          AMD EPYC 8534P 64-Core Processor
12
  Topology                      1 Processor, 4 Cores
13
  Identifier                    AuthenticAMD Family 25 Model 160 Stepping 2
14
  Base Frequency                2.30 GHz
15
  L1 Instruction Cache          64.0 KB x 4
16
  L1 Data Cache                 64.0 KB x 4
17
  L2 Cache                      512 KB x 4
18
  L3 Cache                      16.0 MB
19

20
内存信息
21
  Size                          3.82 GB
22

23
单核测试分数：1210
24
多核测试分数：4356
25
详细结果链接：https://browser.geekbench.com/v5/cpu/22666764
26
可供参考链接：https://browser.geekbench.com/search?k=v5_cpu&q=AMD%20EPYC%208534P

Bench#

1
----------------------------------------------------------------------
2
 CPU Model          : AMD EPYC 8534P 64-Core Processor
3
 CPU Cores          : 4 @ 2296.548 MHz
4
 CPU Cache          : 512 KB
5
 AES-NI             : ✓ Enabled
6
 VM-x/AMD-V         : ✓ Enabled
7
 Total Disk         : 3.9 TB (3.7 GB Used)
8
 Total Mem          : 3.8 GB (318.6 MB Used)
9
 Total Swap         : 512.0 MB (0 Used)
10
 System uptime      : 0 days, 0 hour 4 min
11
 Load average       : 0.42, 0.25, 0.10
12
 OS                 : Debian GNU/Linux 12
13
 Arch               : x86_64 (64 Bit)
14
 Kernel             : 6.1.0-22-amd64
15
 TCP CC             : bbr
16
 Virtualization     : KVM
17
 IPv4/IPv6          : ✓ Online / ✗ Offline
18
 Organization       : AS210644 AEZA INTERNATIONAL LTD
19
 Location           : Amsterdam / NL
20
 Region             : North Holland
21
----------------------------------------------------------------------
22
 I/O Speed(1st run) : 746 MB/s
23
 I/O Speed(2nd run) : 795 MB/s
24
 I/O Speed(3rd run) : 689 MB/s
25
 I/O Speed(average) : 743.3 MB/s
26
----------------------------------------------------------------------
27
 Node Name        Upload Speed      Download Speed      Latency
28
 Speedtest.net    481.71 Mbps       920.67 Mbps         158.50 ms
29
 Los Angeles, US  554.28 Mbps       5374.82 Mbps        144.38 ms
30
 Dallas, US       694.22 Mbps       6234.85 Mbps        117.44 ms
31
 Montreal, CA     569.74 Mbps       868.78 Mbps         81.44 ms
32
 Amsterdam, NL    9297.40 Mbps      9386.14 Mbps        1.43 ms
33
 Mumbai, IN       626.55 Mbps       6085.88 Mbps        134.77 ms
34
 Singapore, SG    118.73 Mbps       34.11 Mbps          183.45 ms
35
 Tokyo, JP        304.91 Mbps       3170.27 Mbps        251.43 ms
36
----------------------------------------------------------------------

FIO Benchmark#

1
read4k-rand: (groupid=0, jobs=16): err= 0: pid=2807: Tue Jul  9 17:05:14 2024
2
  read: IOPS=1008, BW=4172KiB/s (4272kB/s)(245MiB/60013msec)
3
    slat (usec): min=2, max=65500, avg=15852.93, stdev=1848.37
4
    clat (usec): min=3, max=2078.2k, avg=1981070.86, stdev=209967.15
5
     lat (msec): min=5, max=2094, avg=1996.93, stdev=210.00
6
    clat percentiles (msec):
7
     |  1.00th=[  609],  5.00th=[ 1972], 10.00th=[ 1989], 20.00th=[ 2005],
8
     | 30.00th=[ 2005], 40.00th=[ 2022], 50.00th=[ 2022], 60.00th=[ 2022],
9
     | 70.00th=[ 2039], 80.00th=[ 2039], 90.00th=[ 2039], 95.00th=[ 2039],
10
     | 99.00th=[ 2039], 99.50th=[ 2039], 99.90th=[ 2056], 99.95th=[ 2056],
11
     | 99.99th=[ 2072]
12
   bw (  KiB/s): min= 3920, max= 5304, per=96.93%, avg=4044.58, stdev=23.01, samples=1910
13
   iops        : min=  980, max= 1326, avg=1011.10, stdev= 5.75, samples=1910
14
  lat (usec)   : 4=0.01%, 10=0.02%
15
  lat (msec)   : 10=0.01%, 20=0.02%, 50=0.05%, 100=0.08%, 250=0.25%
16
  lat (msec)   : 500=0.43%, 750=0.41%, 1000=0.43%, 2000=23.04%, >=2000=78.62%
17
  cpu          : usr=0.04%, sys=0.18%, ctx=60071, majf=0, minf=0
18
  IO depths    : 1=0.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=100.0%
19
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
20
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.1%
21
     issued rwts: total=60545,0,0,0 short=0,0,0,0 dropped=0,0,0,0
22
     latency   : target=0, window=0, percentile=100.00%, depth=128
23
read64k-seq: (groupid=1, jobs=16): err= 0: pid=2823: Tue Jul  9 17:05:14 2024
24
  read: IOPS=4001, BW=252MiB/s (264MB/s)(14.6GiB/59355msec)
25
    slat (usec): min=5, max=45435, avg=3995.96, stdev=9042.63
26
    clat (usec): min=2, max=527819, avg=505484.84, stdev=29404.10
27
     lat (usec): min=11, max=527829, avg=509479.87, stdev=28011.28
28
    clat percentiles (msec):
29
     |  1.00th=[  477],  5.00th=[  481], 10.00th=[  489], 20.00th=[  506],
30
     | 30.00th=[  510], 40.00th=[  510], 50.00th=[  514], 60.00th=[  514],
31
     | 70.00th=[  514], 80.00th=[  514], 90.00th=[  514], 95.00th=[  514],
32
     | 99.00th=[  514], 99.50th=[  514], 99.90th=[  518], 99.95th=[  523],
33
     | 99.99th=[  527]
34
   bw (  KiB/s): min=245760, max=263581, per=99.25%, avg=256352.89, stdev=405.26, samples=1888
35
   iops        : min= 3840, max= 4112, avg=4002.68, stdev= 6.33, samples=1888
36
  lat (usec)   : 4=0.01%, 20=0.01%, 50=0.01%
37
  lat (msec)   : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.02%, 50=0.05%
38
  lat (msec)   : 100=0.09%, 250=0.25%, 500=15.05%, 750=85.36%
39
  cpu          : usr=0.02%, sys=1.00%, ctx=58284, majf=0, minf=0
40
  IO depths    : 1=0.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=100.0%
41
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
42
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.1%
43
     issued rwts: total=237490,0,0,0 short=0,0,0,0 dropped=0,0,0,0
44
     latency   : target=0, window=0, percentile=100.00%, depth=128
45
write4k-rand: (groupid=2, jobs=16): err= 0: pid=2839: Tue Jul  9 17:05:14 2024
46
  write: IOPS=558, BW=2370KiB/s (2427kB/s)(139MiB/60201msec); 0 zone resets
47
    slat (usec): min=3, max=208109, avg=28673.75, stdev=70865.63
48
    clat (usec): min=2, max=10004k, avg=3489488.75, stdev=2212516.50
49
     lat (usec): min=23, max=10004k, avg=3517422.75, stdev=2228382.22
50
    clat percentiles (msec):
51
     |  1.00th=[  609],  5.00th=[ 1838], 10.00th=[ 1838], 20.00th=[ 2039],
52
     | 30.00th=[ 2232], 40.00th=[ 2232], 50.00th=[ 2433], 60.00th=[ 2668],
53
     | 70.00th=[ 3272], 80.00th=[ 5134], 90.00th=[ 7550], 95.00th=[ 8557],
54
     | 99.00th=[ 9329], 99.50th=[ 9597], 99.90th=[ 9731], 99.95th=[ 9731],
55
     | 99.99th=[ 9866]
56
   bw (  KiB/s): min=  512, max= 5976, per=94.14%, avg=2231.31, stdev=86.74, samples=1909
57
   iops        : min=  128, max= 1494, avg=557.78, stdev=21.68, samples=1909
58
  lat (usec)   : 4=0.01%, 10=0.04%, 20=0.01%, 50=0.01%
59
  lat (msec)   : 250=0.52%, 500=0.48%, 750=0.41%, 1000=0.36%, 2000=11.38%
60
  lat (msec)   : >=2000=92.84%
61
  cpu          : usr=0.00%, sys=0.04%, ctx=4762, majf=0, minf=0
62
  IO depths    : 1=0.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=100.0%
63
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
64
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.1%
65
     issued rwts: total=0,33620,0,0 short=0,0,0,0 dropped=0,0,0,0
66
     latency   : target=0, window=0, percentile=100.00%, depth=128
67
write64k-seq: (groupid=3, jobs=16): err= 0: pid=2855: Tue Jul  9 17:05:14 2024
68
  write: IOPS=112, BW=7528KiB/s (7708kB/s)(445MiB/60501msec); 0 zone resets
69
    slat (usec): min=46, max=615919, avg=141957.90, stdev=121752.39
70
    clat (usec): min=4, max=29011k, avg=14075348.42, stdev=5403032.38
71
     lat (msec): min=203, max=29087, avg=14224.48, stdev=5441.77
72
    clat percentiles (msec):
73
     |  1.00th=[  625],  5.00th=[ 3071], 10.00th=[ 5738], 20.00th=[ 9194],
74
     | 30.00th=[14429], 40.00th=[14697], 50.00th=[14832], 60.00th=[15368],
75
     | 70.00th=[15905], 80.00th=[16576], 90.00th=[17113], 95.00th=[17113],
76
     | 99.00th=[17113], 99.50th=[17113], 99.90th=[17113], 99.95th=[17113],
77
     | 99.99th=[17113]
78
   bw (  KiB/s): min= 2165, max=24570, per=100.00%, avg=8625.15, stdev=272.62, samples=1196
79
   iops        : min=   23, max=  380, avg=132.64, stdev= 4.28, samples=1196
80
  lat (usec)   : 10=0.19%, 20=0.04%
81
  lat (msec)   : 250=0.22%, 500=0.34%, 750=0.31%, 1000=0.38%, 2000=1.68%
82
  lat (msec)   : >=2000=101.43%
83
  cpu          : usr=0.03%, sys=0.03%, ctx=4753, majf=0, minf=0
84
  IO depths    : 1=0.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=2.7%, 32=7.5%, >=64=89.8%
85
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
86
     complete  : 0=0.0%, 4=99.7%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.3%
87
     issued rwts: total=0,6804,0,0 short=0,0,0,0 dropped=0,0,0,0
88
     latency   : target=0, window=0, percentile=100.00%, depth=128
89

90
Run status group 0 (all jobs):
91
   READ: bw=4172KiB/s (4272kB/s), 4172KiB/s-4172KiB/s (4272kB/s-4272kB/s), io=245MiB (256MB), run=60013-60013msec
92

93
Run status group 1 (all jobs):
94
   READ: bw=252MiB/s (264MB/s), 252MiB/s-252MiB/s (264MB/s-264MB/s), io=14.6GiB (15.7GB), run=59355-59355msec
95

96
Run status group 2 (all jobs):
97
  WRITE: bw=2370KiB/s (2427kB/s), 2370KiB/s-2370KiB/s (2427kB/s-2427kB/s), io=139MiB (146MB), run=60201-60201msec
98

99
Run status group 3 (all jobs):
100
  WRITE: bw=7528KiB/s (7708kB/s), 7528KiB/s-7528KiB/s (7708kB/s-7708kB/s), io=445MiB (466MB), run=60501-60501msec
101

102
Disk stats (read/write):
103
  vda: ios=131687/137160, merge=0/3535, ticks=3133156/6436983, in_queue=9573024, util=94.03%

Below is the GPT-generated explanation based on the above fio results:

Based on the provided disk performance benchmark results, we can draw the following conclusions:

read4k-rand group:

IOPS is 1008 with 4172 KiB/s bandwidth; read throughput is relatively stable.

Average read latency is about 1.99 seconds, and 99% of requests complete within 2.07 seconds.

CPU utilization is low, while most I/O depth is at 64, indicating a heavily loaded storage backend.

read64k-seq group:

IOPS is 4001 with 252 MiB/s bandwidth; read throughput is quite high.

Average read latency is about 509.48 ms, and 99% of requests complete within 0.53 seconds.

CPU utilization is low, with most I/O depth at 64, again suggesting a heavily loaded storage backend.

write4k-rand group:

IOPS is 558 with 2370 KiB/s bandwidth; write throughput is relatively stable.

Average write latency is about 3.49 seconds, and 99% of requests complete within 9.73 seconds.

CPU utilization is low, with most I/O depth at 64, indicating the storage system is under significant pressure.

write64k-seq group:

IOPS is 112 with 7528 KiB/s bandwidth; sequential write throughput is decent.

Average write latency is about 14.22 seconds, and 99% of requests complete within 17.11 seconds.

CPU utilization is low, with most I/O depth at 64, which suggests the backend storage is close to saturation.

Overall, the disk performance benchmark shows generally acceptable throughput for both reads and writes, but latency for part of the workload is quite high. Further tuning and optimization of the storage system may be required to improve performance.