Howdy, Stranger!

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


Is this KVM VPS too slow?
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.

Is this KVM VPS too slow?

harrysdtharrysdt Member
edited April 2014 in Reviews

dd if=/dev/zero of=iotest bs=64k count=16k conv=fdatasync && rm -rf iotest
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 108.094 s, 9.9 MB/s

This is a 1024MB KVM VPS, this is very slow I think

«134

Comments

  • I do that after a reboot already....

  • 0xdragon0xdragon Member
    edited April 2014

    That's a little odd. Have you tried contacting their support? :)

  • Question is, do you experience any slowness when you actually use it?

    Thanked by 2netomx Lee
  • MunMun Member

    No, the real question is did he put a ticket in or go to IRC and ask for it to be looked at or did he just come here to complain. If the latter, go make a ticket and stop pissing on this forum with your support tickets. LET is NOT a helpdesk, and your provider doesn't sit here all day twittling his thumbs waiting for a thread like this.

    https://www.bluevm.com/submitticket.php?step=2&deptid=2 <-- fill out info here, (after logging in)

    Thanked by 1rm_
  • FritzFritz Veteran

    I'm wondering where is the location?

  • rm_rm_ IPv6 Advocate, Veteran
    edited April 2014

    serverian said: Question is, do you experience any slowness when you actually use it?

    That's like the fire department saying to a person who screams "help, my house is on f*cking fire!"
    "yes sir, but what is the actual problem? do you feel discomfort from the smoke, or is it just too hot to stay inside?"

  • @rm_ said:
    "yes sir, but what is the actual problem? do you feel discomfort from the smoke, or is it just too hot to stay inside?"

    It's nothing like that. He doesn't even know if that result is slow or not. He asks us if it's slow.

    Thanked by 1netomx
  • ryanarpryanarp Member, Patron Provider

    I wouldn't even try arguing @serverian you can't win and at worse you will be called a enemy of dd tests.

  • rm_rm_ IPv6 Advocate, Veteran
    edited April 2014

    serverian said: It's nothing like that. He doesn't even know if that result is slow or not. He asks us if it's slow.

    9.9 MB/sec at

    said: dd if=/dev/zero of=iotest bs=64k count=16k conv=fdatasync

    is slow, and a matter for intervention by the provider (or service cancellation by the customer in case of a lack thereof). Your question seemed to imply that "if it doesn't feel slow, then everything is fine and dandy", which couldn't be further from the truth. I know you providers actively try to discredit and put a stigma on testing with dd, but this particular set or parameters used is fine, and it does in fact objectively measure about the most important aspect of VPS performance (I/O speed). Regarding the actual numbers, at around 50 it'd be "okay, a bit low, but let's see your ioping", but sub-10MB is way too slow no matter how you look at it.

  • netomxnetomx Moderator, Veteran

    @rm_ said:
    is slow, and a matter for intervention by the provider (or service cancellation by the customer in case of a lack thereof). Your question seemed to imply that "if it doesn't feel slow, then everything is fine and dandy", which couldn't be further from the truth. I know you providers actively try to discredit and put a stigma on testing with dd, but this particular set or parameters used is fine, and it does in fact objectively measure about the most important aspect of VPS performance (I/O speed). Regarding the actual numbers, at around 50 it'd be "okay, a bit low, but let's see your ioping", but sub-10MB is way too slow no matter how you look at it.

    How about OVH vps? They have the same DD output

  • @rm_ unless 20 customers decide to put the same dd test in cron, running at the same time. Then the results can be... interesting.

  • serverianserverian Member
    edited April 2014

    @rm_ said:
    is slow, and a matter for intervention by the provider (or service cancellation by the customer in case of a lack thereof). Your question seemed to imply that "if it doesn't feel slow, then everything is fine and dandy", which couldn't be further from the truth. I know you providers actively try to discredit and put a stigma on testing with dd, but this particular set or parameters used is fine, and it does in fact objectively measure about the most important aspect of VPS performance (I/O speed). Regarding the actual numbers, at around 50 it'd be "okay, a bit low, but let's see your ioping", but sub-10MB is way too slow no matter how you look at it.

    Both dd "tests" and ioping show nothing about the truth.

    Here is the dd "test" of newer Backupsy nodes:

    root@XXXXX:~# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync; unlink test
    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 1.01357 s, 1.1 GB/s

    A storage server is faster than a pure ssd raid10 array?

  • rm_rm_ IPv6 Advocate, Veteran

    netomx said: How about OVH vps? They have the same DD output

    So? It's a bad result no matter which provider. Or do you want to make a point that just because OVH does this, and says "it's normal", the 10MB/sec I/O is the new golden standard of performance? On OVH it's that low due to rate-limiting, but the end result will be the same, a VPS which is performing worse than a different one with, say, 100 MB/sec I/O. The 'dd' test is just the quickest and simplest way to detect and demonstrate this problem.

  • netomxnetomx Moderator, Veteran

    @rm_ said:
    So? It's a bad result no matter which provider. Or do you want to make a point that just because OVH does this, and says "it's normal", the 10MB/sec I/O is the new golden standard of performance? On OVH it's that low due to rate-limiting, but the end result will be the same, a VPS which is performing worse than a different one with, say, 100 MB/sec I/O. The 'dd' test is just the quickest and simplest way to detect and demonstrate this problem.

    Nope. I just want to know why do you need to read 1.1GB lineally more than 10MB/s

  • rm_rm_ IPv6 Advocate, Veteran

    rds100 said: unless 20 customers decide

    Ah... "Everyone is just running 'dd' at the same time", overseller's favourite excuse.

    to put the same dd test in cron, running at the same time.

    Who said anything about putting anything into cron?

    serverian said: A storage server is faster than a pure ssd raid10 array?

    Sure, a very high result can be misleading due to various reasons (e.g. onboard memory on the RAID card), but this proves nothing about the cases when it's very low.

  • @rm_ said:
    Sure, a very high result can be misleading due to various reasons (e.g. onboard memory on the RAID card), but this proves nothing about the cases when it's very low.

    The problem is, everyone says dd is wrong and shouldn't be used to measure performance. However, noone suggests which tool to use to actually benchmark disk performance or if it's even necessary.

    When you first deploy your application, benchmark it with a tool like speedcurve.com and save the results. (Or if it's not a web application, run it and check the load values of your VPS and write them down)

    If you are experiencing slowness on your VPS,

    1. Check your load if it's below the allocated vCPUs
    2. Run a MTR to check if there is packet loss on the network
    3. Run speedcurve.com again and compare the results

    If you really get slow results, you do not have run dd or ioping, you just show the results of both tests to your provider and then your provider can check the node for abuse (if they are not already monitoring it), and run IO benchmark tests themselves to determine if it's an IO problem and solve it.

    You do not have to use primitive benchmarking to see if your VPS is slow or not.

  • rm_rm_ IPv6 Advocate, Veteran
    edited April 2014

    Time is money. Either you can run a quick 'dd' and pinpoint the problem instantly, or you can waste a lot of effort doing elaborate crap, the primary purpose of which is to pander to a provider using a suboptimal or oversold I/O configuration. "So now I have a crap I/O, but hey, at speed curve whatever website it's still not too bad! [because the disk caching mostly manages to paper over the issue for now]"

    Thanked by 2linuxthefish Mark_R
  • rm_rm_ IPv6 Advocate, Veteran
    edited April 2014

    serverian said: When you first deploy your application, benchmark it with a tool like speedcurve.com

    serverian said: you just show the results of both tests to your provider

    And they will be more than happy to respond with "it must be your application", or "our VPS is unmanaged, please kindly optimize your application yourself". dd or ioping on the other hand are the most basic, primitive as you said, raw benchmarks no one can really argue with. And then if a provider still tries to argue (like you do here) that a crap sub-10MB/sec dd is "normal", it's a sure sign to run far far away and never look back.

    Thanked by 1Mark_R
  • @rm_ said:
    And they will be more than happy to respond with "it must be your application", or "our VPS is unmanaged, please kindly optimize your application yourself". dd or ioping on the other hand are the most basic, primitive as you said, raw benchmarks no one can really argue with. And then if a provider still tries to argue (like you do here) that a crap sub-10MB/sec dd is "normal", it's a sure sign to run far far away and never look back.

    You are missing the point here.

    OP does not say his VPS is slow, he's asking if it's slow based on the dd test.

    People were selling VPS on single PATA drives not too long ago. And people were not complaining that their sites are slow. Because most of the web applications use 75-80% reads, not writes.

    10MB/sec of dd test does not necessarily mean the VPS is slow. You think it's slow because LET is turned to a dd pissing contest after serverbear.

  • For clarity, it may be a sign of slowness. But it doesn't mean anything by itself.

    For another clarity, I'm not defending over-loaded servers. Yet, all the VPSes hosted with our company get at least >200MB/sec from that dd "test".

    Thanked by 1Infinity
  • NekkiNekki Veteran

    serverian said: OP does not say his VPS is slow, he's asking if it's slow based on the dd test.

    I don't think he is actually. Ignore the title, the content of the post makes it clear that he knows it's slow.

  • nonubynonuby Member
    edited April 2014

    @Mun said:
    No, the real question is did he put a ticket in or go to IRC and ask for it to be looked at or did he just come here to complain. If the latter, go make a ticket and stop pissing on this forum with your support tickets. LET is NOT a helpdesk, and your provider doesn't sit here all day twittling his thumbs waiting for a thread like this.

    https://www.bluevm.com/submitticket.php?step=2&deptid=2 <-- fill out info here, (after logging in)

    Disagree. Shouldnt a reputable company be proactively monitoring the health of the nodes and performance level of guest VMs. This is pretty similar to "we'll move you to another node" shit. Shouldnt really need to point this out, but then again as long as participants here keep lowering expectations then nothing surprises me..

    Any provider that needs a client to notify them of suboptimal performance is basically a turd in my books

  • @Nekki said:
    I don't think he is actually. Ignore the title, the content of the post makes it clear that he knows it's slow.

    If so, how did he determine it's slow? What are his other findings? That's the question. (again)

  • NekkiNekki Veteran

    @serverian said:
    If so, how did he determine it's slow? What are his other findings? That's the question. (again)

    Why does he need 'other findings' - currently, he has apparently crappy I/O; let's assume that (in the absence of information to the contrary) that he was undertaking some I/O intensive operations that ran like poop, so he ran the DD test and got the results above.

    Just to flip this about a bit; lets say this guy is your customer, and he raises a ticket stating that I/O is slow, and gives you the DD test above; what would you ask him to provide before you'd decide to investigate?

  • You purchased from a lowend host. Expect lowend performance. BlueVM are not top tier, they're lower/mid range providers & their prices clearly reflect that.

    Take that into consideration, and why have you not submitted a ticket regarding this?

  • Why does everything on LowEndTalk blow up?

  • @Nekki said:
    Just to flip this about a bit; lets say this guy is your customer, and he raises a ticket stating that I/O is slow, and gives you the DD test above; what would you ask him to provide before you'd decide to investigate?

    It depends on the provider's target IO speed for their clients really. If the provider is targetting 200MB/sec and the customer sent a ticket and saying 300MB/sec is slow for them, then I'd kindly say that that's over our target speed. If it's below that, I'd remind them that they are in a shared environment and there may be other processes using a lot of IO at that moment and ask him if he checked that earlier and got the same results. If so, then there is something wrong with the node and I'd go investigate. (We do monitor the performance manually and automatically by provisioning a test VM and checking how long it takes to get online)

    But another provider may not have any target benchmark performance for the VMs and look at the real world performance and ask for more information from the client like if untarring taking long or ssh being sloppy, or if their web site taking too long to load, etc.

  • NekkiNekki Veteran

    serverian said: It depends on the provider's target IO speed for their clients really. If the provider is targetting 200MB/sec and the customer sent a ticket and saying 300MB/sec is slow for them, then I'd kindly say that that's over our target

    Buuuuut - how are would you measure I/O to tell if you're meeting your own target?

    @LowEND said:
    Why does everything on LowEndTalk blow up?

    Because we come for the cheap VPS and stay for the drama.

  • @Nekki said:

    Checking the iowait times, monitoring the IOPS, and then running dd.

  • rm_rm_ IPv6 Advocate, Veteran
    edited April 2014

    viCommunications said: You purchased from a lowend host. Expect lowend performance.

    Literally every other host around here will provide you with at least 10-20x higher performance than what we see in the first post, for about the same price as BlueVM, or even cheaper. So please don't even start the b/s line that a low end box must also perform like crap, that's not true and has been disproven by not just the top-rated hosts, but by a majority of them. If you don't see any way to provide a cheap VPS other than via extreme overselling, then sorry but this market is not for you at all. Also what is this "viCommunications" name, is it some provider? Remind me to never buy any service from you.

Sign In or Register to comment.