All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Hostcram i9 2vcpu VPS impressions - 3 months later
As I write this post, three months have passed since I signed up for the HostCram VPS plan
2vcpu,
I9-11900k , 3 GB RAM that was on offer here on LET. My post captures some impressions about the VPS, the service and the host, and finally what’s next. At the time of posting this offer HostCram had offered an additional month for posting reviews. You may have seen many reviews with YABS and other BMs, while the second part of this post does have those numbers (from back in September), it is the first part that I think matters more to me as a user. And I believe some of you by extension.
a. Network
This VPS is in Dallas location, for me in India the speeds I get to this vps aren’t the greatest. That is the bad news, fortunately the only bad news. Depending in your use case, this may or may not be a deal breaker. Leaving results from ‘ping’
PING104.xxx.xxx.xx 56(84) bytes of data. 64 bytes from 104.xxx.xxx.xx: icmp_seq=2 ttl=52 time=253 ms 64 bytes from 104.xxx.xxx.xx: icmp_seq=3 ttl=52 time=257 ms
b. Performance
The horsepower is quite impressive. I out the VPS to use for optimization of a batch of 5,000 jpeg and png images (using jpegoptim and optipng) and then converting them to webp. The batch process (es) took a couple of hours. I ran the same processes in another vps. The second VPS had different specs; and was from a different provider. This was a “control system” - 1 vcpu ryzen 3900/ 1 GB RAM, 40 GB NVME ). Same operations took over 5 hours.
I may be stating the obvious here, that faster machines perform better, but note that these may not be the typical processes one runs every day. Unless one runs and image hosting/optimization service. But I thought this was a good way to kick the tyres.
There was no control panel for this web, so re installing will require you to raise a ticket.
Other than that, the there are little reasons to complain. I opted for Debian 11 U 20.04 during install, and have started with it.
c. Service and communication
Uptime has been great.
Because of my image experimentation, ai ended up using over 90 percent of the available disk space. This resulted in a message from HostCram, offering to move my service to a different node. I had the original images, the optimized images, and the webp versions all on the same disk. Removing the originals and the webp (the latter after transfer to a storage vps) solved the issue.
Note that I was travelling then (as is the case now), and I had asked HostCram for a few days before I acted on their ticket. They were patient enough. Two days later, space was freed up.
d. What’s next:
Practically the 3 month service ends today. I remain un decided on continuing this service, not because this is a bad service but because I have too many idlers. plus am away from home so no point in taking a hasty decision. I do have a 1 GB BF offer from this provider ($7, yay!) so there is continuity though in a different format.
Obligatory benchmarks
Edit: will try and post a more recent YABS if time permits
YABS and bench.monster
-------
Tue Sep 28 14:01:27 UTC 2021
Basic System Information:
---------------------------------
Processor : 11th Gen Intel(R) Core(TM) i9-11900K @ 3.50GHz
CPU cores : 2 @ 3356.526 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM : 4.0 GiB
Swap : 4.0 GiB
Disk : 78.2 GiB
Preparing system for disk tests...bash: line 436: curl: command not found
Fio binary download failed. Running dd test as fallback....
dd Sequential Disk Speed Tests:
---------------------------------
| Test 1 | Test 2 | Test 3 | Avg
| | | |
Write | 2.4 GB/s | 2.5 GB/s | 2.5 GB/s | 2.47 GB/s
Read | 5.0 GB/s | 5.1 GB/s | 5.1 GB/s | 5.07 GB/s
bash: line 633: curl: command not found
iperf3 binary download failed. Skipping iperf network tests...
Tue Sep 28 14:02:40 UTC 2021
Basic System Information:
---------------------------------
Processor : 11th Gen Intel(R) Core(TM) i9-11900K @ 3.50GHz
CPU cores : 2 @ 3416.794 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM : 4.0 GiB
Swap : 4.0 GiB
Disk : 78.2 GiB
fio Disk Speed Tests (Mixed R/W 50/50):
---------------------------------
Block Size | 4k (IOPS) | 64k (IOPS)
------ | --- ---- | ---- ----
Read | 1.05 GB/s (263.9k) | 2.88 GB/s (45.0k)
Write | 1.05 GB/s (264.6k) | 2.89 GB/s (45.2k)
Total | 2.11 GB/s (528.5k) | 5.77 GB/s (90.2k)
| |
Block Size | 512k (IOPS) | 1m (IOPS)
------ | --- ---- | ---- ----
Read | 3.48 GB/s (6.7k) | 1.28 GB/s (1.2k)
Write | 3.66 GB/s (7.1k) | 1.37 GB/s (1.3k)
Total | 7.14 GB/s (13.9k) | 2.65 GB/s (2.5k)
iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider | Location (Link) | Send Speed | Recv Speed
| | |
Clouvider | London, UK (10G) | 791 Mbits/sec | 140 Mbits/sec
Online.net | Paris, FR (10G) | 781 Mbits/sec | 106 Mbits/sec
WorldStream | The Netherlands (10G) | 782 Mbits/sec | 175 Mbits/sec
Biznet | Jakarta, Indonesia (1G) | busy | busy
Clouvider | NYC, NY, US (10G) | 822 Mbits/sec | 501 Mbits/sec
Velocity Online | Tallahassee, FL, US (10G) | 807 Mbits/sec | 743 Mbits/sec
Clouvider | Los Angeles, CA, US (10G) | 818 Mbits/sec | 553 Mbits/sec
Iveloz Telecom | Sao Paulo, BR (2G) | 466 Mbits/sec | 119 Mbits/sec
Geekbench 5 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 1618
Multi Core | 2024
Full Test | https://browser.geekbench.com/v5/cpu/10122915
Bench.monster
---------------------------------------------------------------------------
OS : Ubuntu 20.04.3 LTS (64 Bit)
Virt/Kernel : Lxc / 5.4.140-1-pve
CPU Model : 11th Gen Intel(R) Core(TM) i9-11900K @ 3.50GHz
CPU Cores : 2 @ 4804.291 MHz x86_64 16384 KB Cache
CPU Flags : AES-NI Enabled & VM-x/AMD-V Enabled
Load Average : 2.70, 3.18, 2.93
Total Space : 79G (1.0G ~2% used)
Total RAM : 4096 MB (67 MB + 199 MB Buff in use)
Total SWAP : 4047 MB (0 MB in use)
Uptime : 0 days 0:1
---------------------------------------------------------------------------
ASN & ISP : AS39618, HostCram LLC
Organization : HostCram LLC
Location : Dallas, United States / US
Region : Texas
---------------------------------------------------------------------------
## Geekbench v4 CPU Benchmark:
Single Core : 7312 (EXCELLENT)
Multi Core : 12911
## IO Test
CPU Speed:
bzip2 : 177 MB/s
sha256 : 317 MB/s
md5sum : 880 MB/s
RAM Speed:
Avg. write : 6144.0 MB/s
Avg. read : 11298.1 MB/s
Disk Speed:
1st run : 1.1 GB/s
2nd run : 945 MB/s
3rd run : 918 MB/s
-----------------------
Average : 996.5 MB/s
## India Speedtest.net
Location Upload Download Ping
---------------------------------------------------------------------------
Nearby 729.12 Mbit/s 883.73 Mbit/s 3.542 ms
---------------------------------------------------------------------------
India, New Delhi (Weebo) 41.52 Mbit/s 89.77 Mbit/s 261.516 ms
India, Mumbai (SevenStar) 59.01 Mbit/s 56.01 Mbit/s 234.901 ms
India, Bengaluru (I-ON) 19.01 Mbit/s 13.37 Mbit/s 259.965 ms
India, Nagpur (optbb) 2.96 Mbit/s 99.55 Mbit/s 267.177 ms
India, Patna (Airtel) 29.84 Mbit/s 72.96 Mbit/s 266.481 ms
India, Kolkata (Vodafone) 18.40 Mbit/s 11.94 Mbit/s ping error!
India, Visakhapatnam (Alliance) 41.91 Mbit/s 64.71 Mbit/s 260.373 ms
India, Hyderabad (I-ON) 46.03 Mbit/s 94.88 Mbit/s 259.724 ms
India, Madurai (Niss Broadband) 19.47 Mbit/s 14.75 Mbit/s 256.528 ms
---------------------------------------------------------------------------
Finished in : 13 min 4 sec
Timestamp : 2021-09-29 01:20:25 GMT
Saved in : /root/speedtest.log
Share results:
- https://www.speedtest.net/result/12106716822.png
- https://browser.geekbench.com/v4/cpu/16362753
- https://clbin.com/xm9fD
Comments
Did you set CPU limit somehow?
Storage overselling detected.
@vyas11
Im assuming you got it during a promotion? How much are you paying? Maybe you can transfer it to someone else if you're not interested in keeping it.
(I certainly like the numbers I'm seeing
)
@yoursunny
Surely everyone here knows that every low end provider overallocates resources
Network sure, same with CPU (and RAM if it's LXC or OVZ) but overselling storage really worries me, basically the only thing that would break the VPS if it ran out of resources (LXC VPS would just cap ram, same with CPU and network port, but hard to just tell a VM that it has no more storage when it's partitioned otherwise). Storage is also probably the cheapest component. Thin provisioning is fine, but that doesn't seem to be what this is.
its 2021. storage should be dedicated.
It was more about to the NVMe performance.
Yes. We do oversell storage on some LXC nodes. Specially on the node from we sold $3/quarter LXC-4G.
That's not an issue as we don't host more than 20 VPS usually and always keep the usage around 50% to 70%.
This surprises me even more! If you oversell storage, but still aim to keep each node at 50% usage, presumably the implication of that is that you're overselling my significantly more than 2x.
Did I suggest anywhere that each node at 50 percent?
For everyone's peace of mind > https://imgur.com/a/r0GGCPF
One node has managed to go ahead of our target disk usage.
Many of the VPS on this node won't be renewed and we will see improvement here soon.
Ask some other providers how much they oversell on OpenVZ and LXC nodes then come back to me so you can compare how much I am overselling.
Are you hosting the containers in
/?I don't think there is any issue in overselling disk unless people are getting what they have been promised, further I think LowEnd is only possible when resources are smartly oversold because not everyone uses all the resources at same time.
There is local and local-lvm storage on that node you're talking about.
We put OS and template on local and LXC on local-lvm.
The actual usage: https://i.imgur.com/N1XSeG5.png
Well, in the post I replied to, you said:
Perhaps you meant something else, but a natural reading of this sentence is that you would ideally keep disk usage to around 50%, but don't get worried about it until it hits 70%.
And my point is that if you only need to maintain a safety margin in the case of substantial overselling. If you only oversold 2x, then there's no point keeping disks at 50%, because even if everybody doubled their disk usage, you'd hit 100%, so no problem. The fact that you have a 50% target at all, shows that you are substantially more oversold that that.
No thanks, I've got better things to do with my time than do competitive analysis for you.
@ralf doing is planning is something I am good at.
Your implications are just pointless and not everyone going to double their disk usage over night.
We can always migrate some VPS to other nodes if something like you're talking about ever happens. We always keep some nodes on idle as per my expension and disaster recovery plan.
I have already explained my point of view. I also have better things to do myself and going to ignore future comments on this thread.
Good luck.
Just caught up with some of the discussion around storage space.
If I un intentionally stirred a hornet’s nest, @shakib, my apologies.
On the other hand, the conversation is rather enlightening to say the least. Personally I do not have an issue if host reaches out when disk space hits 90 percent utilisation for long periods of time.
Generally speaking, I thought that some systems run optimally at 80 percent utilisation, beyond 85 percent they actually become efficient. Hotels and airlines come to mind. Spikes over 90 percent are okay occasionally, not for sustained period though.
I do not have enough knowledge about disk storage to understand the matter at hand. But applying similar principle, if use exceeds over 90 percent for sustained periods, maybe some alarm switches go off somewhere??
@vyas11
The consequences of overselling servers will be much much less problematic than airlines/hotels(assuming the Provider is properly monitoring the nodes!), since in the worst case scenario, a spare server can be plugged in and someone can be moved. Hotels/airlines obviously can't do that
On the contrary, it's beneficial as it allows to provide services for cheaper and increase profit margins. The higher the density, the more profitable generally.
And do rest assured that any proper provider has alarms, or at least someone monitoring. And there's definitely one of those in play here, as evidenced by you receiving that relocation offer(which I hope included some kind of compensation, but for the price ranges we see around here I doubt it)
While I am not a provider, I am a sysadmin so I have some idea of how these things work; I apologize for any inaccuracies (and if someone more experienced than I am is able to correct me, please do!)
Wait, do you compress already compressed files (especially with JPEG which is "known" for massive generation loss) and then compressing it again? And on top of that you're compressing with such old compressors?
Any reasons why you are not using more recent multithreaded compressors?
Like this https://github.com/fhanau/Efficient-Compression-Tool
Or this (Windows-only) https://css-ig.net/pingo
Idk if you're unaware of that or these old tools (jpegoptim is 19year old at this point, ofc it was updated but base is still 19 years old...) have some pro's that I'm unaware of.
@AXYZE
Thanks for the info on efficient compression tools.
Fair point.
I had used jpegoptim/optipng to check their efficiency and quality drop for a different purpose.Have done some work on image optimization/ compression and CDNs over the past couple of years.
Similar set of images was tested for size reduction across different images compression tools
Tinypng, shortpixel, image compressor etc.
I ran the optimizations in that vein. There is a body of SEO practitioners (and website speed checker services) that recommend keeping image size as small as possible for faster loading times. While I do not agree with that approach 100 percent, I see merit in keeping lower image size for slower networks.