Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Contabo new raw NVMe speed product line benchmark/review - Page 2
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Contabo new raw NVMe speed product line benchmark/review

2

Comments

  • @dufu said:
    Can someone confirm that nested virtualization (svm cpu flag on AMD) is indeed enabled on the production systems? @jsg mentioned the hypervisor flag is enabled on his test systems, but I’m not sure this is the same as svm and vmx flags?

    Contabo support told me nested virtualization is not possible. Maybe they are misinformed?

    They are not misinformed, contabo doesnt allow it. They also imply IO limit, so your results will be a lot worse than it is presented here.

    Jsg doesnt believe contabo test servers before launch prepared for reviewer differ from what customers get and when presented with real evidence he just said that "script is wrong". I could buy new Contabo server just to prove, but... it was already proven so what the point? His opinion, his review. Im responding to you and thats all.

    I will not give login detail to random person, he wont buy server to check (understandable, no problem with that) so take it for what it is.

    If you decide to buy server then please post cpuinfo, yabs so we can see if contabo gave jsg "unlimited server" or "script is wrong". And seems like their support is also wrong. Yea.

  • Which would you recommend since Contabo has NVMe now, Contabo VPS XL or Netcup RS 8000 G9 in regards to performance?

  • bruh21bruh21 Member, Host Rep

    @emilv said:
    Which would you recommend since Contabo has NVMe now, Contabo VPS XL or Netcup RS 8000 G9 in regards to performance?

    netcup will have dedicated cpu resources, but contabo is cheaper. you pick based on that

  • @emilv said:
    Which would you recommend since Contabo has NVMe now, Contabo VPS XL or Netcup RS 8000 G9 in regards to performance?

    NetCup or PHP-Friends are best if you want good, sustained performance. If you want hourly billing, flexibility etc. then Hetzner. Contabo SSD offer strength was generous space, but with NVMe it's not so great compared to Netcup and Netcup has a very high IOPS limit & 2.5Gbps network so you can really push a lot of data.

    Thanked by 1JasonM
  • jsgjsg Member, Resident Benchmarker

    @bruh21 said:

    @emilv said:
    Which would you recommend since Contabo has NVMe now, Contabo VPS XL or Netcup RS 8000 G9 in regards to performance?

    netcup will have dedicated cpu resources, but contabo is cheaper. you pick based on that

    They both offer VPS with dedicated resources (incl. dedi vCore) and netcup occasionally runs promos whith cheap(er) 'root servers' (netcups term for the thing). I myself have a netcup root server and am happy with it.

  • ArkasArkas Moderator

    What are the differences between 'root servers' and 'V server' in netcup, other than the way they charge? I don't understand their German site, just like php friends. Someone should translate their sites.

  • bruh21bruh21 Member, Host Rep
    edited August 2021

    @Arkas said:
    What are the differences between 'root servers' and 'V server' in netcup, other than the way they charge? I don't understand their German site, just like php friends. Someone should translate their sites.

    http://netcup.com

    and the root server has dedicated cpu while the vserver is shared like usual VPS

    Thanked by 1Arkas
  • ArkasArkas Moderator

    @bruh21 said: and the root server has dedicated cpu while the vserver is shared like usual VPS

    Thanks, that does sound good!

  • AXYZEAXYZE Member
    edited August 2021

    @Arkas said:
    What are the differences between 'root servers' and 'V server' in netcup, other than the way they charge? I don't understand their German site, just like php friends. Someone should translate their sites.

    netcup.de is in german shows yearly pricing https://www.netcup.de/vserver/
    netcup.eu is in english shows monthly pricing https://www.netcup.eu/vserver/
    It doesn't matter on which site you order, its same control panel with english.

    Root server has dedicated thread just for you so it means they wont throttle you if you push server to its limits and that why I consider this offer a lot more cost effective. Also they include 2.5Gbps and normal "Vserver" has 1Gbps and lower traffic limit.

    I wouldn't really call it dedicated CPU as you still get penalty depending on second thread of core which belongs to someone else, but this will be veeery small. I get 0.1% max steal on Netcup root servers. It's very minor thing, but I want to point it out so you wont compare it 1:1 with dedicated server or something. I would call it dedicated vCPU, resources are dedicated but its still virtual.

    If you want you can use 5euro off coupon: 36nc16281461204
    Its affcode with bonus for you like for DigitalOcean etc.
    You can redeem it here https://www.netcup.de/bestellen/gutschein_einloesen.php

  • @AXYZE said: If you decide to buy server then please post cpuinfo, yabs so we can see if contabo gave jsg "unlimited server" or "script is wrong". And seems like their support is also wrong. Yea.

    yabs has correctly implemented the VM-X check, so not sure how @jsg claims that the script is wrong.
    https://github.com/masonr/yet-another-bench-script/blob/master/yabs.sh#L187

    It's also somewhat shady that @jsg refuses to even release the binaries for his benchmark.

  • jsgjsg Member, Resident Benchmarker

    @stevewatson301 said:
    It's also somewhat shady that @jsg refuses to even release the binaries for his benchmark.

    I don't refuse. In fact, I published the link long ago.

  • @jsg said:

    @stevewatson301 said:
    It's also somewhat shady that @jsg refuses to even release the binaries for his benchmark.

    I don't refuse. In fact, I published the link long ago.

    Can you please provide a link here? I and many others would like to independently confirm your findings and benchmark other systems.

    Thanked by 2adly agentwp
  • DPDP Administrator, The Domain Guy

    @stevewatson301 said:

    @jsg said:

    @stevewatson301 said:
    It's also somewhat shady that @jsg refuses to even release the binaries for his benchmark.

    I don't refuse. In fact, I published the link long ago.

    Can you please provide a link here? I and many others would like to independently confirm your findings and benchmark other systems.

    https://disk.yandex.com/d/GS1rPud83WKtWs

  • jsgjsg Member, Resident Benchmarker

    @DP said:

    @stevewatson301 said:

    @jsg said:

    @stevewatson301 said:
    It's also somewhat shady that @jsg refuses to even release the binaries for his benchmark.

    I don't refuse. In fact, I published the link long ago.

    Can you please provide a link here? I and many others would like to independently confirm your findings and benchmark other systems.

    https://disk.yandex.com/d/GS1rPud83WKtWs

    That's v.1 but maybe still useful for some.

    The link to the current (and much enhanced) version is here -> https://disk.yandex.com/d/yfbjWW1Oudar6w

  • Amazing. Thanks.

    v.1 in Pascal and v.2 in Nim. Interesting!

  • jsgjsg Member, Resident Benchmarker

    As I was asked by someone to benchmark his Contabo VPS I came across a point I meant to mention but then forgot; now I was reminded:

    Contabo's - like many providers using template based installs - VPSs only come with a 'root' account (plus password), so many customers just stick with that. Don't! Always create - and use - a normal user account! and then disable root login in the SSHD config!

    Thanked by 1Kassem
  • @jsg said:
    As I was asked by someone to benchmark his Contabo VPS I came across a point I meant to mention but then forgot; now I was reminded:

    Contabo's - like many providers using template based installs - VPSs only come with a 'root' account (plus password), so many customers just stick with that. Don't! Always create - and use - a normal user account! and then disable root login in the SSHD config!

    You should acquire a free Oracle VM running the Epyc CPU's and not the ARM ones. You'll discover something interesting when you know they are disk and cpu limited.

  • @TimboJones said:

    @jsg said:
    As I was asked by someone to benchmark his Contabo VPS I came across a point I meant to mention but then forgot; now I was reminded:

    Contabo's - like many providers using template based installs - VPSs only come with a 'root' account (plus password), so many customers just stick with that. Don't! Always create - and use - a normal user account! and then disable root login in the SSHD config!

    You should acquire a free Oracle VM running the Epyc CPU's and not the ARM ones. You'll discover something interesting when you know they are disk and cpu limited.

    Do you want me to give him the bad news?

    Thanked by 2adly drunkendog
  • @stevewatson301 said:

    @TimboJones said:

    @jsg said:
    As I was asked by someone to benchmark his Contabo VPS I came across a point I meant to mention but then forgot; now I was reminded:

    Contabo's - like many providers using template based installs - VPSs only come with a 'root' account (plus password), so many customers just stick with that. Don't! Always create - and use - a normal user account! and then disable root login in the SSHD config!

    You should acquire a free Oracle VM running the Epyc CPU's and not the ARM ones. You'll discover something interesting when you know they are disk and cpu limited.

    Do you want me to give him the bad news?

    He needs to see it for himself.

  • I moved from a contabo VPS L SSD to a VPS L NVMe (had to set it from scratch because Contabo's rescaling capabilities are non-existent) and the latter is ridiculously slow in terms of storage access. Even ran a comparison with a much lower specs hetzner VPS and the difference is night to day

    What's going on, what am I missing?

    In terms of software running, I setup ubuntu 20.04, hardened it using standard procedures and I am running some apps that also run with other providers. Like I said, I turned everything off and ran that comparison and got really surprised at how crappy my results were

    What am I missing? Have I been lied to?

  • @mindanalyzer said:
    I moved from a contabo VPS L SSD to a VPS L NVMe (had to set it from scratch because Contabo's rescaling capabilities are non-existent) and the latter is ridiculously slow in terms of storage access. Even ran a comparison with a much lower specs hetzner VPS and the difference is night to day

    What's going on, what am I missing?

    In terms of software running, I setup ubuntu 20.04, hardened it using standard procedures and I am running some apps that also run with other providers. Like I said, I turned everything off and ran that comparison and got really surprised at how crappy my results were

    What am I missing? Have I been lied to?

    Youre missing nothing, normal case with Contabo.

    Stick with Hetzner/Contabo/PHP-Friends if you want quality on budget. Better performance with same money, no setup fees (so its even cheaper), working DDoS Protection etc.

    Buy Contabo only if you need SSD on cheap. NVMe plan is not cost effective, you will get more, faster storage with netcup.

  • @mindanalyzer said:
    I moved from a contabo VPS L SSD to a VPS L NVMe (had to set it from scratch because Contabo's rescaling capabilities are non-existent) and the latter is ridiculously slow in terms of storage access. Even ran a comparison with a much lower specs hetzner VPS and the difference is night to day

    What's going on, what am I missing?

    In terms of software running, I setup ubuntu 20.04, hardened it using standard procedures and I am running some apps that also run with other providers. Like I said, I turned everything off and ran that comparison and got really surprised at how crappy my results were

    What am I missing? Have I been lied to?

    wait for the server review king to explain it in layman terms to you.

    Thanked by 1adly
  • ArkasArkas Moderator

    Get whatever is best for you. For production, I would stick with Hetzner. Best Bang for your buck IMO

  • @mindanalyzer said:
    I moved from a contabo VPS L SSD to a VPS L NVMe (had to set it from scratch because Contabo's rescaling capabilities are non-existent) and the latter is ridiculously slow in terms of storage access. Even ran a comparison with a much lower specs hetzner VPS and the difference is night to day

    What's going on, what am I missing?

    In terms of software running, I setup ubuntu 20.04, hardened it using standard procedures and I am running some apps that also run with other providers. Like I said, I turned everything off and ran that comparison and got really surprised at how crappy my results were

    What am I missing? Have I been lied to?

    Contabo is heavily oversold, and you are getting what you pay for, since they are so cheap. I wouldn't put anything performance sensitive on contabo.

    @cybertech said:

    wait for the server review king to explain it in layman terms to you.

    The server king can't handle any criticism, he will blame the benchmark script, or that this user doesn't understand how a VPS works. Only his closed source benchmark script is correct and everyone else is wrong. And then he will get hostile that some LET users hate him and are attacking him. We don't need anymore of that.

  • FalzoFalzo Member
    edited September 2021

    just because you buy a VPS with NVMe doesn't mean you see better performance. it keeps being a shared environment, so NVMe only means the provider can cram even more customers onto the same poor array while keeping the IO numbers and limits they had before.

    it's also a good marketing term, but not neccessarily beneficial to clients and especially some like contabo are known to force their nodes and sell ressources really to the limits in every regard.

    that's not bad per se... just might be needed for offering at that dirt cheapo price like they do.

  • jarjar Patron Provider, Top Host, Veteran
    edited September 2021

    I'm personally really enjoying my Docker server on Contabo. I don't really do benchmarks anymore, I've found that I'm much more satisfied by merely asking "Is this doing what I want? Am I happy with it?" It's running the NextCloud instance that houses all of my nudes, doing great. Also my first time really getting into Docker Compose, and damn does it make things nice and portable. It's an SSD VPS from literally hours before the NVMe ones released. Not seeing any need to change.

    Thanked by 2Falzo adly
  • jsgjsg Member, Resident Benchmarker
    edited September 2021

    I'm probably going to disappoint some, but I have to largely agree with those who criticize Contabo's VMs.

    Background story/info/explanation:

    The numbers in my review were honest and correct And my review reflected what 100+ benchmark runs had measured, minus what seemed to be a realistic and educated estimation of empty node vs. occupied node.
    I can say that because I had access to the test VPS for a while after sales started and I still benchmarked to get a good real-world based impression of how speed would decline after sales start (which is normal but should stay within a reasonable frame). As you might know I also log the time of each benchmark run, so I could actually get a pretty clear picture of how far and how fast the performance numbers of the diverse sub-systems (processor, disk IO, ...) went down as node occupation increased.

    Some days ago I changed/swapped one of my own private VPSs with Contabo - as a normal "everybody" customer - from SSD to NVMe, and restarted benchmarking. Those current "just any VPS in between thousands of VPSs" benchmarks clearly showed very significantly decreased disk performance.

    Well noted, as already said above, I did also keep testing the test-VPS after sales started (among other reasons because I wanted to be sure that my guesstimate actually was realistic) and I indeed kept seeing numbers going down but still staying above or about what I had predicted and from everything I can tell, in the last days of my testing the node was full and basically a normal production node.

    So, how come that now, just a short while later, my benchmark shows considerably worse results on a normal production node?
    Obviously the first assumption coming to mind is that Contabo played dirty and gave me a special "golden" node for my benchmarking. I can however exclude that explanation with a very high degree of confidence - except for one factor: I do presume that they kept an eye on my node that is, if there was any user crossing his limits they'd highly likely reign in on him rather quickly (or move him to another node, or ...). Basically it comes down to "no provider wants to have a problematic node benchmarked and reviewed. Of bloody course even squeaky clean honest providers want a tested node not being one with, say heavy abusers on it.

    Maybe those like, to name one example @AXYZE who say that Contabo harshly enforces (supposedly low) limits are right. I don't know; I myself do not experience unreasonably low limits but then my benchmark behaves well (remember: I explicitly designed it to be a good neighbour), but I can not and do not want to exclude the possibility that those critics are right.

    My personal guess is that there are many abusers on Contabo VPS, 'abuser' meaning a user who frequently or even consistently uses considerably more than his fair share -and- that Contabo has some weaknesses in their node and VM monitoring. Certain other (partially weak) points in the software area seem to suggest that the software side is not their strong point ...

    So, does that mean that Contabo VPSs are crap? Nope, in the (relatively few) tests I ran so far on my NVMe VPS I see quite good performance, I'm a happy camper and feel that I definitely get my moneys worth - but I do not see an exceptionally great VM anymore. Disk IO for example still is in the range of the fastest I've see so far, but to be honest and clear it is far below what my first benchmark run showed; example: where I first saw 600/s MB I now see about 270 MB/s. That's still a very nice result and in the ballpark of the best 5 I've seen so far anywhere but still, that's less than half of what I saw earlier.

    Accordingly I will ask Contabo to remove my quote from their site.

    I've said it already but I will repeat it: My first loyalty is with the LET community and not with any provider. And my integrity is not for sale.

    Thanked by 1seenu
  • @jsg said:
    My personal guess is that there are many abusers on Contabo VPS, 'abuser' meaning a user who frequently or even consistently uses considerably more than his fair share -and- that Contabo has some weaknesses in their node and VM monitoring.

    Does running hundred bench for several days continuously categorized as abuse activity?

  • good we all agree on contabo now!

  • jsgjsg Member, Resident Benchmarker

    @chocolateshirt said:

    @jsg said:
    My personal guess is that there are many abusers on Contabo VPS, 'abuser' meaning a user who frequently or even consistently uses considerably more than his fair share -and- that Contabo has some weaknesses in their node and VM monitoring.

    Does running hundred bench for several days continuously categorized as abuse activity?

    With my benchmark probably not, but still, I've seen even test VMs given explicitly for testing generating "you have reached you traffic limit. Please contact support" emails.
    While vpsbench of course also can go hardcore I activate that mode only rarely, e.g. when testing a dedi, or for a short time, e.g. when probing caches.
    Without the "good neighbour" mode my benchmarking highly likely would trigger abuse flags though.

Sign In or Register to comment.