Howdy, Stranger!

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


Bad I/O Performance on Crissic VPS
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.

Bad I/O Performance on Crissic VPS

damandaman Member

Hi everyone,
Few weeks ago, I purchased a 512MB VPS from Crissic.net and It was working pretty well. I ran some benchmark and found that the I/O performance was around 150mb/s so I contacted their support regarding this issue. They told me that this is a pretty good value. I was hosted on Jax03 node. After 1-2 weeks, I again ran a benchmark and found that I/O performance was worst (around 50mb/s or sometimes even more lower). So again I contacted their support regarding this issue. After sometime they moved my account to Jax08 node and I was now getting 220mb/s I/O performance. But today, when I'm running I/O test using dd command, it shows speed around 56.2MB/s.

Here is the report:

username@server:~$ sudo dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync 16384+0 records in 16384+0 records out 1073741824 bytes (1.1 GB) copied, 19.1114 s, 56.2 MB/s

I don't know think, this is a good value. Tell me guys what can be done. Is it a good value?

Comments

  • sc754sc754 Member

    said: support regarding this issue. They told me that this is a pretty good value. I was hosted on Jax03 node.

    If it's not ssd cached / ssd then this would be normal. It just means some other people on the disk were using it at the same time as you...

  • You're being an abusive client.

    What's more important to test (if you must test something) is the latency of your I/O requests, not the total throughput available. A bare hardware RAID10 on a dedicated server (with a premium RAID card) and four hard drives; only has about 500MB/sec TOTAL POSSIBLE THROUGHPUT.

    It's important to note that the RAID card's cache is what allows the 500MB/sec when a regular hard drive can only do about 70MB/sec theoretical maximum.


    You can test the latency with the ioping command, which should return less than 50ms on a regular hard drive, and much less on a hardware raid configuration (but not in all cases, such as worst case scenarios (outer border, on-copy or out-of-cache reads.)

    Thanked by 1Maximum_VPS
  • Could be better but if they say that's ok then there's nothing you can do except move provider.

  • I remember they advertised 600MB/s?

  • FrankZFrankZ Veteran

    What @GoodHosting said. Don't be fooled by cache performance. If you are over 30MB/s on a non SSD node don't worry about it too much as real world performance should be ok.

  • GoodHosting said: when a regular hard drive can only do about 70MB/sec theoretical maximum.

    my Seagate hard drive will do nearly 180MB/S and its for home usage so that comment is utter nonsense.

  • I believe for some reason they limit I/O for about 100 MB/s

  • Ian_Ian_ Member

    But really what are you using the vps for? Have you tried the dd tests at different times.

    Thanked by 1TheHackBox
  • damandaman Member

    Ian_ said: But really what are you using the vps for? Have you tried the dd tests at different times.

    I'm using this VPS for hosting a WordPress blog (nginx, php-fpm, MariaDB) with no traffic. Yes, I tried using it different times. Sometimes I get good I/O but sometimes it is worst.

  • damandaman Member

    Noerman said: I believe for some reason they limit I/O for about 100 MB/s

    But I don't think 56MB/s is a good value.

  • @daman said:
    But I don't think 56MB/s is a good value.

    Agreed.

  • damandaman Member

    GoodHosting said: You're being an abusive client.

    What do you mean? I did not understood.

  • I am a customer of crissic. i also have complain about this and open some ticket, and donald from crissic reply "We have not had time to test the IO scheduling deeply. I do not have the answer to that question unfortunately. The idea is that you'll get the I/O that you need.". So it is what it is. My lowest IO are about 20-ish mbit/sec but on normal average about 100ish mbit/sec. When i first sign up, the IO are 300mbit/sec or better. Overall i like crissic, the support are good and the server are stable, and i am happy customer for now. But, if there are better offer and no improvement from crissic, i am willing to move next year (buy yearly plan). Ping @SkylarM

  • FrankZFrankZ Veteran
    edited March 2014

    @Noerman said:
    I believe for some reason they limit I/O for about 100 MB/s

    I don't think so. It's shared I/O and depends on what your neighbors on the node are doing.



    @daman @Noerman - If this is soooo bad, how was everyone was able to manage a website with IDE drives a few years ago, when a 4 drive RAID 10 would max out at 85MB/s. Just saying that we have become spoiled by hardware performance advances and some people no longer tune their setup for performance and just expect the provider to overcome their system design shortcomings. This is LET afterall



    Good news is Pure SSD solves this issue and there is a special here
    http://lowendtalk.com/discussion/23972/ftpit-ssd-sale-cheap-offers-inside-yearly-specials#latest

    Thanked by 2TarZZ92 Maximum_VPS
  • MaouniqueMaounique Host Rep, Veteran

    sirmbhe said: The idea is that you'll get the I/O that you need.

    That is more or less so. I find that 30 MB/s and stable ioping is good enough for almost anything. If you need intensive disk usage, such as a lot of queries on a DB which cannot be cached, then you go for SSD, preferably full SSD array.

    Thanked by 1ironhide
  • sirmbhesirmbhe Member
    edited March 2014

    @Maounique said:
    That is more or less so. I find that 30 MB/s and stable ioping is good enough for almost anything. If you need intensive disk usage, such as a lot of queries on a DB which cannot be cached, then you go for SSD, preferably full SSD array.

    This is right, cause i do not need blazing fast IO speed. And as i say, i am happy with crissic. Skylar and his team support are very nice (quick and helping), and for me IO are not critical. But if i can get more speed, i won't refuse it :-D
    Recently i just purchase a 128MB plan ssd cached from ramnode, and the speed are just too sweet to see (and feel lol). i never have ssd plan, now i wonder how it will be? :D (fast of course, stupid question)

  • MaouniqueMaounique Host Rep, Veteran

    sirmbhe said: now i wonder how will it be? :D

    Exactly the same as those tests are run in RAM.

  • SkylarMSkylarM Member
    edited March 2014

    Not sure where you're pulling ~50 or so, we cannot replicate that. If it's really that low that would be an issue, but pulling 150 like in your original ticket is NOT an issue. We explained that in great detail in the ticket. We do run larger disk arrays (12 drives in RAID10) however there is also a rather large handful of customers per node as well, which will eat at the overall result of a "DD" test.

    (node itself)
    [root@jax08 vz]# 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, 3.72272 s, 288 MB/s

    (VPS on the same node)
    [root@vps /]# 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, 4.22718 s, 254 MB/s

    Your consistency on a test may vary from time to time, but that's to be expected with how much these drives are being used. Real world performance, however, should not be impacted.

    @sirmbhe said:

    Running 20MB/s or so is way below what any of our newer nodes should be running. 300MB/s sounds like you had been provisioned on a fairly empty node, so some degrading is going to happen over time. I'd strongly suggest opening a ticket if you see regular results below 100. Anything over 100 isn't really anything that I'd be super worried about.

  • sirmbhesirmbhe Member
    edited March 2014

    @SkylarM said:
    Running 20MB/s or so is way below what any of our newer nodes should be running. 300MB/s sounds like you had been provisioned on a fairly empty node, so some degrading is going to happen over time. I'd strongly suggest opening a ticket if you see regular results below 100. Anything over 100 isn't really anything that I'd be super worried about.

    Hello, @SkylarM. i have already open a ticket, remember? if not, see ticket #183537 i put some bench result in there. Or should i open new ticket? :-)

  • @sirmbhe said:
    Hello, SkylarM. i have already open a ticket, remember? if not, see ticket #183537 i put some bench result in there. Or should i open new ticket? :-)

    Can you update the ticket with any issue that is still current and as much detail as you can provide? I'm not currently showing any outstanding issues with that node.

  • @SkylarM said:
    Can you update the ticket with any issue that is still current and as much detail as you can provide? I'm not currently showing any outstanding issues with that node.

    I remember you advertised 600MB/s?

  • @trexos said:
    I remember you advertised 600MB/s?

    On an empty or near empty node yeah. We've since modified our adverts to be less misleading.

  • @daman, how long do you need to complete # apt-get install nginx?

  • VDS6VDS6 Member
    edited March 2014

    Pretty good speed for lowend vps. Not sure why people expect lowend to be fast. I personally stand for limiting io for lowend servers.

    On one hand, I understand that some clients do not like it, but others do not mind it and you still get a share of a market.

    But on the other hand if you do not limit io, i remember a client who got pissed simply because his server did not have ipv4 IP, which was listed on the website, and he did the following after loosing paypal claim: He made a crontab with dd command to ran every single 2 minutes just to harm others on the node.

    Since than i personally stand for limiting io for low cost services that do not make much of profit.
    But it's simply my IMHO.

  • MaouniqueMaounique Host Rep, Veteran
    edited March 2014

    VDS6 said: But on the other hand if you do not limit io, i remember a client who got pissed simply because his server did not have ipv4 IP, which was listed on the website, and he did the following after loosing paypal claim: He made a crontab with dd command to ran every single 2 minutes just to harm others on the node.

    That is why people which file disputes have to be suspended/terminated immediately, regardless of the outcome.

    Thanked by 1linuxthefish
  • HostNunHostNun Member
    edited March 2014

    @Liam said:
    What are you running? 56MB/s is not that bad for 99% of usage.

    This.

    The question "is it a good value?" is meaningless unless you relate it to your own specific needs. For instance, is there any reason that whatever it is you're developing 'needs' more than 56MB/s? Do you even notice a difference? Do you even lift?

    And what about the tortoise and the hare, huh? Huh?

  • edited March 2014

    A big difference, when we can setup our fresh installed os for 10minutes from other provider, and we took about more than 1 hours in here.

    Most indonesian users only use Crissic for tunnelling, aka free internet bypassing. So they only need ssh, not a webserver.

    Thanked by 1sirmbhe
Sign In or Register to comment.