[rudra]

Review of VPS 4

Generously Sponsored by Shadow Hosting

after 7 months of uses and abuses :-P

---

Today is 31st January, 2020 and I got this crazy VPS from our awesome Post4VPS giveaway in the first week of July, 2019. It is a KVM VPS located in Germany ( Hetzner Online gmbh Datacentre ).

Let me get the specs out there,

Specifications:

Processor: Intel Xeon E5-1650 v2
Topology: 2 Processors, 2 Cores [from Geekbench 5]

Code: (Select All)

root@shadow:~# cat /proc/cpuinfo
...
model name: Intel(R) Xeon(R) CPU E5-1650 v2 @ 3.50GHz
...
cpu MHz: 3499.990
cache size: 12288 KB
...
bugs: cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf mds swapgs itlb_multihit

bogomips: 6999.98

bogomips is actually 2 * 6999.8 . Cause that was per thread.
That bug-details really worried me. But ! .. yes ! But the following result will put you at ease !

Code: (Select All)

root@shadow:~# lscpu
Architecture:                    x86_64
CPU op-mode(s):                  32-bit, 64-bit
Byte Order:                      Little Endian
Address sizes:                   46 bits physical, 48 bits virtual
CPU(s):                          2
On-line CPU(s) list:             0,1
Thread(s) per core:              1
Core(s) per socket:              1
Socket(s):                       2
NUMA node(s):                    1
Vendor ID:                       GenuineIntel
CPU family:                      6
Model:                           62
Model name:                      Intel(R) Xeon(R) CPU E5-1650 v2 @ 3.50GHz
Stepping:                        4
CPU MHz:                         3499.990
BogoMIPS:                        6999.98
Hypervisor vendor:               KVM
Virtualization type:             full
L1d cache:                       32 KiB
L1i cache:                       32 KiB
L2 cache:                        256 KiB
L3 cache:                        12 MiB
NUMA node0 CPU(s):               0,1
Vulnerability Itlb multihit:     KVM: Vulnerable
Vulnerability L1tf:              Mitigation; PTE Inversion
Vulnerability Mds:               Mitigation; Clear CPU buffers; SMT Host state unknown
Vulnerability Meltdown:          Mitigation; PTI
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:        Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:        Mitigation; Full generic retpoline, IBPB conditional, IBRS_FW, STIBP disabled, RSB filling
Vulnerability Tsx async abort:   Not affected

So clearly most of them are mitigated ! Which is an awesome news !!

So that means 2 cores with a total of 2 threads.
RAM: 10GB

Code: (Select All)

root@shadow:~# cat /proc/meminfo
MemTotal:       10205088 kB

Code: (Select All)

root@shadow:~# lshw | grep ecc
capabilities: ecc
configuration: errordetection=multi-bit-ecc

Storage: HDD 300 GB
Network: 1 Gbps
Bandwidth: unmetered (FUP)
Virtualization: KVM - full hardware virtualisation
IP address: 1 IPv4
Control Panel: Virtualizor
Location: Germany

I upgraded to latest dev version of Ubuntu during all this, which is 20.04

Code: (Select All)

root@shadow:~# cat /etc/os-release
NAME="Ubuntu"
VERSION="20.04 LTS (Focal Fossa)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu Focal Fossa (development branch)"
VERSION_ID="20.04"

along with the latest mainline kernel, manually installed.

Code: (Select All)

root@shadow:~# uname -a
Linux shadow 5.5.0-050500-generic #202001262030 SMP Mon Jan 27 01:33:36 UTC 2020 x86_64 x86_64 x86_64 GNU/LinuX

---

VPS Usage:

I hope none will mind if I only mention the regular things I do on this VPS. I do not want to embarrass myself.

>> Downloading and working through different versions of source codes to find the changes and learn and modify them. I have very slow internet and I move a lot. So this server is like my online lab that is always there for me.

>> Learning to configure old and new softwares and set ups on VPS. I do all kind of weird combinations with that stuff. It helps me learn things a lot.

>> I was going to host some websites. two of those I moved away to a much lower spec VPS and one friend who had loads of pdf educational materials and a fairly big site with good traffic went ahead and bought a multi-year plan from a very useless webhost just to save money and now suffering as hell.

>> I have developed multiple websites on it. From fairly simple ones to multi-page adaptive ones.

>> I am really afraid to use too much of CPU and hurt the other users on the machine besides me. So I never run compiles on it, which I hope I will get to do when I buy a paid one from them when I am ready and have some more time to spare.

>> We all lesser mortals are in the shadows of HR here. Though I have no complaint. He deserves that. I hope I will find some time soon to write up some of the crazy stuffs I learned to do while using it.

>> I will probably be taking part in an online data science course and I am sure this VPS will be a godsend. AI has been a constant source of excitement to me. Though I am more interested in the theories.

---

Performance:

I will start with cpu benchmarks. Then it will be storage and network in that order.

CPU Benchmarks -

I actually did three runs of Geekbench 5. It uses i3 8100 as base with 1000 score and these scores are relative to that. So as you can see, with two threads, it is like having two cores of i3 8100, almost, with ECC RAM. will talk more on this in the final words at the end.*1*
First one with Ubuntu 19.04, kernel 5.0.0-36-generic
718 - Single-Core Score
1432 - Multi-Core Score
Now on Ubuntu 20.10 with kernel 5.4.0-12-generic
725 - Single-Core Score
1387 - Multi-Core Score

This is Ubuntu 20.10 with kernel 5.5.0-050500-generic

729 - Single-Core Score
1459 - Multi-Core Score

Here is a search at Geekbench 5 browser with i3 8100.
i3 8100

I won't be doing Unixbench. Cause I think they are old enough to be even more useless than geekbench. Given all the specialised ISA and hardware accelerators, it is getting really hard to compare general performance in a meaningful way. So it is always better to know exactly what you want to do and find out the various possible ways of implementing them and may be then you could go on comparing their performance on various parameters. That will be a heavily specialised one. But much more useful than a general score thrown out by such benchmark softwares.

Storage Benchmark:

dd:

NOTE - Everyone loves to see dd results on reviews. But keep it in mind that dd is single threaded and sequential. So it is one of the worst in giving you an idea on real life performance of your storage. Especially on a VPS that involves shared environment and multiple threads.
   fio and ioping are the proper utilities for that. But since testing with fio and a 4GB will take 50 minutes or more, I will refrain from doing that in this case. If you really want to see the results, you may request in your reply below and I will comply.

write speed -

Code: (Select All)

1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 23.7881 s, 45.1 MB/s
root@shadow:~# sync; dd if=/dev/zero of=tempfile bs=1M count=1024; sync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 27.7186 s, 38.7 MB/s

Doing 10 times and taking the average, it comes at 44.3MB/s, a pretty good score on HDD. I can confirm that I have seen it go higher, but never below 35MB/s.

Read speed -
Read speed from cache,

Code: (Select All)

root@shadow:~# dd if=tempfile of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 0.133126 s, 8.1 GB/s
root@shadow:~# dd if=tempfile of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 0.167328 s, 6.4 GB/s

Doing 10 times and taking average, I got 7.9 GB/s.

Direct read speed from HDD -

Code: (Select All)

root@shadow:~# /sbin/sysctl -w vm.drop_caches=3
vm.drop_caches = 3
root@shadow:~# dd if=tempfile of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 23.7659 s, 45.2 MB/s

Reapeating it 10 times and taking average, the result is 51.2 MB/s. I must say I am happy with this.

hdparm - cached

Code: (Select All)

root@shadow:~# hdparm -Tt /dev/sda
/dev/sda:
Timing cached reads:   22106 MB in  1.99 seconds = 11110.30 MB/sec
Timing buffered disk reads: 254 MB in  3.01 seconds =  84.46 MB/sec
root@shadow:~# hdparm -Tt /dev/sda

/dev/sda:
Timing cached reads:   21948 MB in  1.99 seconds = 11029.98 MB/sec
Timing buffered disk reads: 380 MB in  3.00 seconds = 126.60 MB/sec
root@shadow:~# hdparm -Tt /dev/sda

/dev/sda:
Timing cached reads:   21660 MB in  1.99 seconds = 10885.05 MB/sec
Timing buffered disk reads: 284 MB in  3.00 seconds =  94.66 MB/sec

hdparm - direct

Code: (Select All)

root@shadow:~/S# hdparm -Tt --direct /dev/sda                                  
/dev/sda:
Timing O_DIRECT cached reads:   576 MB in  2.03 seconds = 284.31 MB/sec
Timing O_DIRECT disk reads: 258 MB in  3.01 seconds =  85.69 MB/sec
root@shadow:~/S# hdparm -Tt --direct /dev/sda

/dev/sda:
Timing O_DIRECT cached reads:   728 MB in  2.00 seconds = 363.66 MB/sec
Timing O_DIRECT disk reads: 294 MB in  3.01 seconds =  97.68 MB/sec
root@shadow:~/S# hdparm -Tt --direct /dev/sda

/dev/sda:
Timing O_DIRECT cached reads:   838 MB in  2.00 seconds = 418.36 MB/sec
Timing O_DIRECT disk reads: 314 MB in  3.01 seconds = 104.34 MB/sec

Using S - this project on github aims to mimic real life usage to calculate performance. Here we see a throughput test under various read and write loads using different schedulers.
    This is a much better way to do the testing. It uses fio and results are said to be more realistic.

Code: (Select All)

root@shadow:~/S# ./run_multiple_benchmarks/run_main_benchmarks.sh "throughput"

Computing overall stats
-------------------------------------------------------
# throughput
# First column: Workload
# Next columns: Aggregate throughput [MB/sec], or X if results are
#               unreliable because workloads did not stop when asked to
# Reference case: 1r-seq
# Reference-case meaning: Peak rate with one sequential reader
#
# Workload           mq-deadline                 bfq               kyber        
       none
 10r-rand               0.379               0.554            0.517714          
 0.434857
 10r-seq              37.8521             39.6158             37.8905          
  30.2475
 1r-seq               55.2812             61.4688             45.9338          
   57.625
 5r5w-rand                0.71               0.696               0.763        
  0.962333
 5r5w-seq              35.716             34.1175             33.7495          
    28.49
-------------------------------------------------------

I recently installed graphical interface (gui). You could get some idea of how fast the system responds from its bootup time.

Code: (Select All)

root@shadow:~# systemd-analyze
Startup finished in 2.008s (kernel) + 34.846s (userspace) = 36.854s
graphical.target reached after 34.830s in userspace

Awesome timing on an HDD system !

Finally we come to ioping.
ioping
disk seek rate -

Code: (Select All)

root@shadow:~/S# ioping -R /dev/sda                                            
--- /dev/sda (block device 300 GiB) ioping statistics ---
600 requests completed in 3.01 s, 2.34 MiB read, 199 iops, 798.3 KiB/s
generated 601 requests in 3.01 s, 2.35 MiB, 199 iops, 798.6 KiB/s
min/avg/max/mdev = 257.6 us / 5.01 ms / 95.7 ms / 10.0 ms

disk sequential speed (MiB/s)

Code: (Select All)

root@shadow:~/S# ioping -RL /dev/sda

--- /dev/sda (block device 300 GiB) ioping statistics ---
1.45 k requests completed in 2.96 s, 361.8 MiB read, 489 iops, 122.4 MiB/s
generated 1.45 k requests in 3.00 s, 362 MiB, 482 iops, 120.7 MiB/s
min/avg/max/mdev = 416.0 us / 2.04 ms / 111.2 ms / 6.62 ms

Disk I/O latency in various modes:

Default mode

Code: (Select All)

root@shadow:~/S# ioping -c 20 /tmp

--- /tmp (ext4 /dev/sda1) ioping statistics ---
19 requests completed in 67.2 ms, 76 KiB read, 282 iops, 1.10 MiB/s
generated 20 requests in 19.0 s, 80 KiB, 1 iops, 4.21 KiB/s
min/avg/max/mdev = 495.3 us / 3.54 ms / 17.1 ms / 5.45 ms

Asynchronous mode

Code: (Select All)

root@shadow:~/S# ioping -c 20 -A /tmp
--- /tmp (ext4 /dev/sda1) ioping statistics ---
19 requests completed in 88.6 ms, 76 KiB read, 214 iops, 858.2 KiB/s
generated 20 requests in 19.0 s, 80 KiB, 1 iops, 4.21 KiB/s
min/avg/max/mdev = 506.5 us / 4.66 ms / 31.6 ms / 7.56 ms

Direct mode

Code: (Select All)

root@shadow:~/S# ioping -c 20 -D /tmp
--- /tmp (ext4 /dev/sda1) ioping statistics ---
19 requests completed in 98.3 ms, 76 KiB read, 193 iops, 773.3 KiB/s
generated 20 requests in 19.0 s, 80 KiB, 1 iops, 4.21 KiB/s
min/avg/max/mdev = 493.7 us / 5.17 ms / 27.2 ms / 6.58 ms

Cached mode

Code: (Select All)

root@shadow:~/S# ioping -c 20 -C /tmp
--- /tmp (ext4 /dev/sda1) ioping statistics ---
19 requests completed in 201.6 us, 76 KiB read, 94.2 k iops, 368.1 MiB/s
generated 20 requests in 19.0 s, 80 KiB, 1 iops, 4.21 KiB/s
min/avg/max/mdev = 6.89 us / 10.6 us / 12.4 us / 1.08 us

Write mode

Code: (Select All)

root@shadow:~/S# ioping -c 20 -W /tmp
--- /tmp (ext4 /dev/sda1) ioping statistics ---
19 requests completed in 892.3 ms, 76 KiB written, 21 iops, 85.2 KiB/s
generated 20 requests in 19.1 s, 80 KiB, 1 iops, 4.19 KiB/s
min/avg/max/mdev = 6.95 ms / 47.0 ms / 238.0 ms / 51.1 ms

Alternate Read/Write mode(ping-pong)

Code: (Select All)

root@shadow:~/S# ioping -c 20 -G /tmp
--- /tmp (ext4 /dev/sda1) ioping statistics ---
19 requests completed in 619.2 ms, 76 KiB, 30 iops, 122.7 KiB/s
generated 20 requests in 19.0 s, 80 KiB, 1 iops, 4.21 KiB/s
min/avg/max/mdev = 509.0 us / 32.6 ms / 154.6 ms / 36.3 ms

 #comment on ioping results#
     This being a QEMU virtualized disk, the results are consistent with that bit of information. the improvement at cached mode is due to in system caching and the increase in delay at write and alternate read/write mode , which can be considered as more realistic measures, are witness to the effect of virtualization and caching at other levels.

Network Benchmark:
Using the post4VPS benchmark script [thanks to HR and sohamb03]

Code: (Select All)

Speedtest (IPv4 only)
---------------------
Your public IPv4 is ****

LocationProviderSpeed
CDNCachefly50.7MB/s

Atlanta, GA, USColoat13.2MB/s
Dallas, TX, USSoftlayer11.9MB/s
Seattle, WA, USSoftlayer10.2MB/s
San Jose, CA, USSoftlayer9.46MB/s
Washington, DC, USLeaseweb 15.1MB/s

Sao Paulo, Brazil       Softlayer7.34MB/s

Singapore Softlayer9.32MB/s
Taiwan                  Hinet           6.27MB/s
Tokyo, Japan Linode6.52MB/s

Nuremberg, GermanyHetzner97.5MB/s
Rotterdam, Netherlandsid3.net81.5MB/s
Haarlem, NetherlandsLeaseweb102MB/s
Milan, ItalySoftlayer81.5MB/s

Melbourne, AUSoftlayer6.27MB/s

Info for the uninitiated. Nothing to worry about those low figures. The speeds depend on the throughput of the other side too. Here is the proof for the Gbps link.

speedtest.net

Code: (Select All)

root@shadow:~/S# speedtest --share
Retrieving speedtest.net configuration...
Testing from Hetzner Online GmbH (****)...
Retrieving speedtest.net server list...
Selecting best server based on ping...
Hosted by GameAddicted (Lauterbach) [74.17 km]: 3.121 ms
Testing download speed................................................................................
Download: 920.98 Mbit/s
Testing upload speed................................................................................................
Upload: 924.12 Mbit/s

As good as promised !!

SPECIAL NOTE ON HARDWARE:-
 If you run hardware detection utilities, you will see that there is VGA hardware. But do not turn on the VGA option in control panel. It will just keep your system from booting. I had to do a support request for that. Simply because I never expected that to cause this.
  That VGA is an emulated one by QEMU. hence dont expect hardware graphics acceleration.

Final Words :-

I think all the information I have already provided up there speaks for itself. The processor is Ivy Bridge 130W 22nm 6core/12threads chip released in 2013.

*1*
1) You get AVX on it, but no AVX 2, F16C, FMA3 which you get in i3 8100. That is one of the reason why single core  Geekbench 5 score for the i3 is mostly around 950, while for this xeon it hovers around 700. But if you check their average Geekbench 4 scores, they appear much closer in single core rating (both hanging around 4000 with i3 8100 almost always just couple hundred ahead). This Xeon also has almost 1.5 times memory bandwidth, Higher on-chip cache and the benefit of ECC relative to i3 8100, even though it takes DDR3 RAM. So that and the 6c/12t explains its high multi-core performance.
   So basically we have two fairly powerful cores here. More than enough for all the jobs I usually do on it.

1A) Shadow Hosting have very generous CPU usage restriction in their TOS. It says, "On VPS, you may not use 90% CPU for more than 30 minutes.". I have seen 5/10/15 minutes.

2) The storage performance I think is totally awesome for HDD based system. No two words about it.

3) When it comes to network, it brings a smile on my face. Ipv6 allocation would be a nice bonus. But there is HE tunnels, for all your ipv6 needs.

4) no TSX on it. So no hardware bugs from it. Unfortunately the SGX is there. But from our investigations above, it seems most of the issues are well taken care of. Do I need to remind you that it is at Hetzner DC after all. 

Therefore, Considering all the above, I would say Shadow Hosting has a pretty solid platform there and I would not hesitate to recommend them to anyone looking for entry to mid level VPS with enough RAM and storage.

Quality of support:
  First, dear reader, let me remind you that I am on a fairly high spec VPS sponsored by Shadow Hosting which I am using free of cost, as a perk of being an active member of this vibrant community. Hence I never expect [and I think no one at my position here should expect] the level of service that you will get when you are a paying customer. That is a no-brainer. 
    But what surprised me about Shadow Hosting and Manal, the guy running this show, is how quick he is at responding to any question or problem, even here on Post4VPS. So far I had to open two support request threads. The reason for one of them were mentioned above [VGA selection in control panel]. On both occasions, he responded within 12 hours with the solution. It was in the second month of my use. I was doing a lot of experiments back then on ways to set up the VPS. 

In short, The level of service is very prompt and of the highest quality I have ever experienced. I wish it stays this way, so I can keep on recommending them to people in need of a reliable hosting.

Note.
Well, I never ever misuse things ! But "abuse" is a word that can be applied to my use cases sometimes, without much of a stretch to its meaning. I always liked to explore. "How else can I do this ?" or  "Could I really not do it if I tried harder ?" are some of my favourite questions ! And this amazing VPS 4 has never let me down so far !

sadly i got banned for trying to download a tv show torrent with very low seeder count [old unpopular one. not downloading in india].

So everyone looking for VPS from shadow hosting, remember that they actively monitor and ban you for using torrent to download copyrighted material or tv shows, i.e. they don't practice 'looking the other way' like many other providers out there. So very good vps for other uses. But certainly not for torrent use, not even occasionally... not even once..

They don't shy away from being strict, which might actually be in your favour by keeping the resources more available.

[Honey]

Nicely written review, the specs really are amazing if you consider that It is provided free of charge, Internet speeds are pretty standard too.

I'd actually like to ask you to try a Geekbench 3 benchmark on it, I did on VPS 5 in my review and I'd like to see how they both compare

[rudra]

@Honey
You got a sweet deal there. You have dedicated cores. Thats awesome ! Here is the Geekbench result you asked for.
http://browser.primatelabs.com/geekbench3/8841769

I had to get 32 bit support libraries for it. aarrggg

I just noticed your memory performance results. Your single and multicore scores are so close by. the effect of 2 vs 4 channels I think.

This Xeon wins at multi-core performance. Other than that, they are pretty much head to head.

[jagathks]

@rudra, this is such a long detailed nicely written review. Your disk speed is decent for HDD and your network speed is really awesome.
Regarding your usage, no one can mind anything unless you are hurting you neighbor, breaching TOS..

Hope I get some soon for testing..

[Manal]

Let me make you a bit more happy, @rudra.

For the fact, the cores assigned to you are dedicated as far as you don't make it 24/7 and disturb other neighbors. That 90% for 30 minutes is for noisy neighbors and is applicable for precautions only.

If you use too much, and any report of poor CPU comes to me, I'll have to throttle it. Otherwise, technically, there has been no restrictions imposed. Otherwise, there's alot of room to live freely.

[rudra]

Wow ! Thanks Manal ! I surely am a lot happier now !

I think the only way it may happen is if I run a big compile job, like doing kernel or say chrome. I am not going to do that. Another way might be a malware doing something I can't think of. crypto ? Well, I will keep doing my best to keep them away.

But I have a question @Manal.

What do you think about VPS users setting up Gentoo from scratch ? It is a lot of load for a long time. I guess no one wants / allows that kind of load on their VPS.

[tryp4vps]

Very cool and well presented review. Lots of useful data are included.

Overall you have set a new standard of writing high-quality VPS review in this forum. And especially thanks for mentioning the issue of the VGA hardware emulated by Qemu. Didn't know about it.

[Sn1F3rt]

Yet another excellent review from one of the top notch VPS users of this forum. I must say @rudra that this one is a great review of any VPS. I hours VPS 4 got the respect it deserves and ofc you deserved the VPS to make it great. I'm sure @Manal would be extremely happy with this review.

Good to see that you're happy with your VPS. I'm amazed by the I/O and the disk latency even that it's HDD. Moreover, the network speed is great too. Truly VPS 4 is a great VPS and courtesy of @Manal, he's made it even greater when he declared the extra 2GB RAM being added to the VPS, a couple of months ago. (P.S.: My VPS 9 Atlanta is still the best one )

This one's a greatly detailed review and I must say one which all hosting providers expect when they sponsor a VPS. True to my words, it's really amazing!

Regards,

[Manal]

Note.
Well, I never ever misuse things ! But "abuse" is a word that can be applied to my use cases sometimes, without much of a stretch to its meaning. I always liked to explore. "How else can I do this ?" or  "Could I really not do it if I tried harder ?" are some of my favourite questions ! And this amazing VPS 4 has never let me down so far !

I was going through your review and I didn't read much last time. In your case, testing stuff is totally permittable as long as it doesn't violate our policy and those "high-end" testing isn't in production use. You may test as much as you want without any suspension. Just leave some breathing room for other servers in the node . As long it doesn't bother other people in the apartment, you may do whatever you want(within ToS).

This applies to both VPS 4 and VPS 5. However, VPS 5 will need to be a bit moderated because it's not for sale anymore(full with sufficient space and power left behind during mission-critical operations by our customer without bothering other people). VPS 4 can be tested as long as it doesn't violate our ToS. Otherwise, it's all good.

[rudra]

Thanks Manal,

I am always very conscious about it all and do my best not to bother or cause trouble for the other tenants. I also do my best to keep things guarded and patched.

You see I got to use other OS than Linux on VPS 8 and never got any wrap on the knuckles from cubedata. Cause I like to be proactive and responsible in such cases.

I would like to say 'Thank you !' to you and to this awesome community for this opportunity. Thanks a lot !