New on LowEndTalk? Please Register and read our Community Rules.
VPS ran outta juice, nerfed after a month - HostHatch (16GB)

Following is the YABS done right after getting the server (day 1), everything was well and performant:
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
# Yet-Another-Bench-Script #
# v2023-04-23 #
# https://github.com/masonr/yet-another-bench-script #
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
Mon Jun 26 22:51:52 UTC 2023
Basic System Information:
---------------------------------
Uptime : 0 days, 0 hours, 10 minutes
Processor : AMD EPYC 7542 32-Core Processor
CPU cores : 4 @ 2899.998 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM : 15.6 GiB
Swap : 0.0 KiB
Disk : 68.7 GiB
Distro : Debian GNU/Linux 12 (bookworm)
Kernel : 6.1.0-9-amd64
VM Type : KVM
IPv4/IPv6 : ✔ Online / ❌ Offline
IPv4 Network Information:
---------------------------------
ISP : HostHatch
ASN : AS63473 HostHatch, LLC
Host : HostHatch, LLC
Location : Canberra, Australian Capital Territory (ACT)
Country : Australia
fio Disk Speed Tests (Mixed R/W 50/50):
---------------------------------
Block Size | 4k (IOPS) | 64k (IOPS)
------ | --- ---- | ---- ----
Read | 229.87 MB/s (57.4k) | 2.79 GB/s (43.6k)
Write | 230.48 MB/s (57.6k) | 2.80 GB/s (43.8k)
Total | 460.35 MB/s (115.0k) | 5.59 GB/s (87.4k)
| |
Block Size | 512k (IOPS) | 1m (IOPS)
------ | --- ---- | ---- ----
Read | 3.02 GB/s (5.9k) | 4.89 GB/s (4.7k)
Write | 3.18 GB/s (6.2k) | 5.21 GB/s (5.0k)
Total | 6.21 GB/s (12.1k) | 10.10 GB/s (9.8k)
iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider | Location (Link) | Send Speed | Recv Speed | Ping
----- | ----- | ---- | ---- | ----
Clouvider | London, UK (10G) | 1.06 Gbits/sec | 827 Mbits/sec | 156 ms
Scaleway | Paris, FR (10G) | 1.17 Gbits/sec | 896 Mbits/sec | 169 ms
NovoServe | North Holland, NL (40G) | 1.10 Gbits/sec | 1.12 Gbits/sec | 156 ms
Uztelecom | Tashkent, UZ (10G) | 1.65 Gbits/sec | 1.22 Gbits/sec | 128 ms
Clouvider | NYC, NY, US (10G) | 753 Mbits/sec | 516 Mbits/sec | 209 ms
Clouvider | Dallas, TX, US (10G) | 578 Mbits/sec | 621 Mbits/sec | 257 ms
Clouvider | Los Angeles, CA, US (10G) | 953 Mbits/sec | 810 Mbits/sec | 172 ms
Geekbench 6 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 1356
Multi Core | 4396
Full Test | https://browser.geekbench.com/v6/cpu/1732763
And here's the YABS after near month
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
# Yet-Another-Bench-Script #
# v2023-04-23 #
# https://github.com/masonr/yet-another-bench-script #
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
Mon Jul 24 05:24:36 UTC 2023
Basic System Information:
---------------------------------
Uptime : 5 days, 19 hours, 58 minutes
Processor : AMD EPYC 7542 32-Core Processor
CPU cores : 4 @ 2899.998 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM : 15.6 GiB
Swap : 0.0 KiB
Disk : 68.7 GiB
Distro : Debian GNU/Linux 12 (bookworm)
Kernel : 6.1.0-9-amd64
VM Type : KVM
IPv4/IPv6 : ✔ Online / ❌ Offline
IPv4 Network Information:
---------------------------------
ISP : HostHatch
ASN : AS63473 HostHatch, LLC
Host : HostHatch, LLC
Location : Singapore, North West (03)
Country : Singapore
fio Disk Speed Tests (Mixed R/W 50/50):
---------------------------------
Block Size | 4k (IOPS) | 64k (IOPS)
------ | --- ---- | ---- ----
Read | 77.69 MB/s (19.4k) | 642.99 MB/s (10.0k)
Write | 77.89 MB/s (19.4k) | 646.37 MB/s (10.0k)
Total | 155.58 MB/s (38.8k) | 1.28 GB/s (20.1k)
| |
Block Size | 512k (IOPS) | 1m (IOPS)
------ | --- ---- | ---- ----
Read | 2.07 GB/s (4.0k) | 2.15 GB/s (2.1k)
Write | 2.18 GB/s (4.2k) | 2.29 GB/s (2.2k)
Total | 4.26 GB/s (8.3k) | 4.44 GB/s (4.3k)
iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider | Location (Link) | Send Speed | Recv Speed | Ping
----- | ----- | ---- | ---- | ----
Clouvider | London, UK (10G) | 629 Mbits/sec | 575 Mbits/sec | 171 ms
Scaleway | Paris, FR (10G) | 838 Mbits/sec | busy | 166 ms
NovoServe | North Holland, NL (40G) | 575 Mbits/sec | 995 Mbits/sec | 156 ms
Uztelecom | Tashkent, UZ (10G) | 1.34 Gbits/sec | 405 Mbits/sec | 120 ms
Clouvider | NYC, NY, US (10G) | 562 Mbits/sec | 122 Mbits/sec | 210 ms
Clouvider | Dallas, TX, US (10G) | 183 Mbits/sec | 233 Mbits/sec | 268 ms
Clouvider | Los Angeles, CA, US (10G) | 531 Mbits/sec | 278 Mbits/sec | 173 ms
Geekbench 6 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 829
Multi Core | 2034
Full Test | https://browser.geekbench.com/v6/cpu/2018081
Also according to their VPS plan, it comes with75GB
of disk space, and what we got is just 68GB
only.
Is it normal by such vps providers or is there something we missing?
The server is just empty, its not even running any real world applications yet.
Comments
Oversell magic
Hosthatch also oversells?
Noticed people here recommending them with positive reviews.
All VPS offers (expect ones with dedicated cores) are overselled. However someone oversells slightly and someones does it as hard as possible without caring about performance.
Check if there is any steal going on. Try reporting it to support.
Are those YABS from the same VPS?
Location is different from other vps
It may be due to GeoIP databases got update - location is determined from IP address
Well there is CPU cores usage, even when the server is supposed to be idle.
Other servers we used didn't even flinch above
0%
in same idle state.Also wondering what's the deal about disk space not 75GB according to the plan?
Yes from the same VPS.
Better check using top/htop and fdisk utilities
That we also noticed in starting, after purchase we ran 1st YABS and saw the location says "Australia' which is not what we signed up for, it should've been "Singapore".
But now when we ran 2nd YABS after month, the location is changed, disk performance nerfed, network speed garbage.. totally different from what it was on day 1
@jobayer looks like you got a company here mate.
You have been hosthatched
GB <-> GiB
The disk space is most likely due to your disk format and some file reservations. I have some VMs with like 100GB that only 98GB can be used, and still, when I'm at like 92GB the disk is "full"
htop:

fdisk:

That's some high steal % holy
Is the VPS idling at the moment you ran YABS?
try
df -hT --total
It's not really big problem, because YABS is using ip-api.com Free API , maybe it's not yet updated at the time you ran YABS.
Wow!
Other server providers we used had no such fancy stuff.
The numbers were always exact according to the plan, so if the plan said 100GB then it was 100GB available for use.
Yes, there's nothing running on the server so it has to be on idle state.
Network speed can be totally understandable, but how come the Disk speed tests also totally different compared to the 1st YABS, also that Geekbench score is almost half.
Are they bad? Because people here suggested others about Hosthatch in positive way.
There has been several reports of HostHatch's VPS performance going down after a few days / months. Check: https://lowendtalk.com/discussion/comment/3692091#Comment_3692091
I've not personally used them though.
Courtesy tag @hosthatch
Interesting. I thought their latest sale in SG would be on an EPYC Milan node, but apparently it's Rome according to the yabs above. My VPS in SG from the sale in March is on EPYC 7443.
Day 1 yabs
Today yabs. I skipped the network test to save bandwidth.
You could be in the oversold node or hypervisor. Ask HostHatch to migrate like my friend here @jobayer
Recent htop activity
Its supposed to be idle, running literally nothing on it so far.
And they say it comes with 16GB ram + 4 AMD EPYC cores
(2 dedicated, 2 fair-shared cores)
So they doing it on purpose to lure people in and silently nerfing the performance of resources after a week or so.
This is pure scam isn't it? Or its like a norm for the Server Providers nowdays.
Overselling CPU cores is the norm. It is a shared environment. However, to what degree depends a lot on the provider. Around 40% decrease in performance should not happen though. I'd suggest you try and benchmark it at other times / another day to see if this is due to a momentary increase on server load or if it is permanently oversold.
They probably just oversell, and if the user doesn't catch up (or he doesn't care) they'll leave it as is. But that user asked for a fix and HostHatch transfered to a different node for him and seems like it fixed his CPU steal.
Either way, this varies a lot depending on the provider. I have some VPS for like $1.20 USD / month which are not oversold in any way, and performance is great.
This is another VM that costs around 1.5€ / month with Ryzen 5 CPU

I have over 30 VMs and I've never seen that CPU steal in my life.
Even if it's a shared environment, a fair-allocation is not the same as overselling directly.
His package has 2 dedicated cores, and I really doubt that's the case.