Howdy, Stranger!

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


Contabo introducing VPS with NVMe drives - Page 5
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 introducing VPS with NVMe drives

12357

Comments

  • @adly said:

    @dev_vps said:
    Well, they are the one putting this quote on their website
    “This NVMe is three times faster than any other NVMe tested (at other providers)”

    That does put the onus on them to justify that quote with some benchmark numbers

    Well, I can say my penis is three times longer than any other penis measured. But I also don't have to show you my penis.

    My analogy would be -

    This highway has 6 lanes (that is more than other highways) but the max only three lanes would be open. And customers driving the most, will be restricted to driving in one lane with 25 miles/hr speed limit.

  • jsgjsg Member, Resident Benchmarker
    edited August 2021

    @dev_vps said:
    The key question that I was referring is this one

    is there a throttle for max I/O that applies to ALL the customers ?

    So far no reply from @contabo_m

    Don't hold your breath. AFAIK they login quite rarely here.

    @adly said:
    @dev_vps that's on you to judge. Contabo is extremely unlikely to admit to anything.

    Which provider is likely to admit any concrete info on what they consider sensitive?

    @dev_vps said:

    @adly said:
    @dev_vps that's on you to judge. Contabo is extremely unlikely to admit to anything.

    Well, they are the one putting this quote on their website
    “This NVMe is three times faster than any other NVMe tested (at other providers)”

    That does put the onus on them to justify that quote with some benchmark numbers

    For a layman customer - does it mean writing 20gb data file (using their VPS with the latest NVMe) will take 1/3 of the time taken for any competitor provider VPS with NVMe storage?

    No, it means that all other factors being equal or very similar that NVMe is 3 times faster than those of competitors (benchmarked so far). And as best I can tell that's still true.

    @dev_vps said:
    My analogy would be -

    This highway has 6 lanes (that is more than other highways) but the max only three lanes would be open. And customers driving the most, will be restricted to driving in one lane with 25 miles/hr speed limit.

    AFAIK not really. They - like pretty much every VPS provider - limit the resources available to a customer (incl. IO rates). As long as customer only slightly or occasionally crosses those boundaries nothing happens. But when some customers (usually called "abusers") frequently and/or massively cross those borders they are limited.
    That's not uncommon but the way most providers seems to handle it.

  • jsgjsg Member, Resident Benchmarker
    edited August 2021

    @dedicatserver_ro said:

    @jsg said: I'm not here to be tested - and certainly not by you who again and again offers tech mumbo jombo to distract from a simple fact: your disks aren't great performers.

    • thats is all ? after I give you all the required information you don't know anymore ?

    Yes, that's all. And that's already way more than your mumbo jumbo distraction attempts merit.

    @jsg said: As for the disks the results are somewhat mixed and in part even disappointing. The reason seems to be that some kind of "Cloud disks" are used (along with a large cache). So the buffered NVMe results are really nice but the direct/sync results are poor and largely about at the level of a crappy SSD.

    • professional response....do you know what you have tested ? - crappy SSD - NVMe Samsung PM1725b

    It's simple. I tested the disks you provided with the VPS purchased from you.

  • @jsg said:

    @dev_vps said:

    @adly said:
    @dev_vps that's on you to judge. Contabo is extremely unlikely to admit to anything.

    Well, they are the one putting this quote on their website
    “This NVMe is three times faster than any other NVMe tested (at other providers)”

    That does put the onus on them to justify that quote with some benchmark numbers

    For a layman customer - does it mean writing 20gb data file (using their VPS with the latest NVMe) will take 1/3 of the time taken for any competitor provider VPS with NVMe storage?

    No, it means that all other factors being equal or very similar that NVMe is 3 times faster than those of competitors (benchmarked so far). And as best I can tell that's still true.

    If I order a VPS from them, would you be willing to run those benchmark tests on my VPS to confirm that my VPS NVMe benchmark is similar to the one you tested with ?

  • jsgjsg Member, Resident Benchmarker

    @dev_vps said:
    If I order a VPS from them, would you be willing to run those benchmark tests on my VPS to confirm that my VPS NVMe benchmark is similar to the one you tested with ?

    You seem to be a decent person, so: Yes, I would.

  • bulbasaurbulbasaur Member
    edited August 2021

    @jsg said: Fact is that gettimeofday() is massively more expensive and slower than clock_gettime()

    The only problem is fio uses neither of these and uses the CPU clocksource unless explicitly overridden. From the man page:

    clocksource=str
    Use the given clocksource as the base of timing. The supported options are:
    gettimeofday gettimeofday(2)
    clock_gettime clock_gettime(2)
    cpu Internal CPU clock source
    cpu is the preferred clocksource if it is reliable

    @jsg said: on some exotic architecture

    x86 is a very real architecture, Spectre is a very real vulnerability, and KPTI is a very real mitigation. There, try again.

    @jsg said: cycles used by the tests (e.g. writing to the disk)

    Can you explain why it's making clock_gettime() calls in a CPU test? (Again, in your attempt to prove me wrong, you overlooked the fact that I was referring to CPU tests):

    [pid  2526] write(1, "----- Processor and Memory -----"..., 33) = 33
    [pid  2526] openat(AT_FDCWD, "/dev/urandom", O_RDONLY) = 3
    [pid  2526] fstat(3, {st_mode=S_IFCHR|0666, st_rdev=makedev(0x1, 0x9), ...}) = 0
    (...)
    [pid  2526] clock_gettime(CLOCK_MONOTONIC, tv_sec=507, tv_nsec=699476566) = 0
    [pid  2526] clock_gettime(CLOCK_MONOTONIC, tv_sec=507, tv_nsec=755654374) = 0
    [pid  2526] clock_nanosleep(CLOCK_REALTIME, 0, tv_sec=0, tv_nsec=2000000, 0x...) = 0
    [pid  2526] clock_gettime(CLOCK_MONOTONIC, tv_sec=507, tv_nsec=757840875) = 0
    [pid  2526] clock_gettime(CLOCK_MONOTONIC, tv_sec=507, tv_nsec=813309269) = 0
    [pid  2526] clock_nanosleep(CLOCK_REALTIME, 0, tv_sec=0, tv_nsec=2000000, 0x...) = 0
    
  • @jsg said:

    @dev_vps said:
    If I order a VPS from them, would you be willing to run those benchmark tests on my VPS to confirm that my VPS NVMe benchmark is similar to the one you tested with ?

    You seem to be a decent person, so: Yes, I would.

    Thank you so much, it is greatly appreciated.

    What configuration would you suggest for this test VPS? So that difference in benchmarks tests can not be attributed to the vps configuration

  • jsgjsg Member, Resident Benchmarker
    edited August 2021

    You'll probably misinterpret that as a victory but I won't discuss with you anymore.

    I'll be nice and answer one question though:

    @stevewatson301 said:
    Can you explain why it's making clock_gettime() calls in a CPU test?

    Measuring the time each test slice takes, duh.

  • jsgjsg Member, Resident Benchmarker

    @dev_vps said:

    @jsg said:

    @dev_vps said:
    If I order a VPS from them, would you be willing to run those benchmark tests on my VPS to confirm that my VPS NVMe benchmark is similar to the one you tested with ?

    You seem to be a decent person, so: Yes, I would.

    Thank you so much, it is greatly appreciated.

    What configuration would you suggest for this test VPS? So that difference in benchmarks tests can not be attributed to the vps configuration

    I'd suggest the configuration that suits your needs. The processor is the same anyway and memory will make some difference but it won't turn a slow NVME into a fast one or vice versa. And after all I guess you want to know what kind of performance you can expect for your needs.

    You are welcome.

  • dedicatserver_rodedicatserver_ro Member, Host Rep

    @jsg said: It's simple. I tested the disks you provided with the VPS purchased from you.

    Therefore you do not know what you are testing or how, you reinvented the wheel and don't even test the same type / configuration of VM , when you are taken into account the results from other tests for example YABS I contradict you you're looking for all kinds of excuses.

    You get from the test:

    • false results ;)
    • wrong expectations 👍
    • wrong storage classifications 👏
  • dedicatserver_rodedicatserver_ro Member, Host Rep
    edited August 2021

    @jsg said: The processor is the same anyway and memory will make some difference but it won't turn a slow NVME into a fast one or vice versa. And after all I guess you want to know what kind of performance you can expect for your needs.

    • NVMe have direct PCIe acces to CPU ( 48 line Intel vs 128 line AMD Epyc , or PCIe 3 vs PCIe4), therefore CPU has a colossal influence on an NVMe interne storage
    • in the tests we did even the CPU frequency between 2.4 GHz silver and 3.0 GHz Gold increased the speed by about 15%
  • jsgjsg Member, Resident Benchmarker
    edited August 2021

    @dedicatserver_ro said:

    @jsg said: It's simple. I tested the disks you provided with the VPS purchased from you.

    Therefore you do not know what you are testing or how, you reinvented the wheel and don't even test the same type / configuration of VM , when you are taken into account the results from other tests for example YABS I contradict you you're looking for all kinds of excuses.

    You get from the test:

    • false results ;)
    • wrong expectations 👍
    • wrong storage classifications 👏

    Uhm, I had NO expectations at all. I simply measured what you provided. Some things I liked, some things - like your disk - were found lacking.
    And I did not classify your storage at all in my review. In fact I do not even care. I simply benchmark the drives I get with a VPS.

    As for (allegedly contradicting) yabs (and other scripts) I don't care. Simple reason: they, or more precisely their lacking, were the reason why I wrote my benchmark in the first place.

    And anyway, I provide reviews to the community that are based on my benchmark. If anyone prefers yabs or whatever benchmarks (s)he is free to ignore mine. Simple.

  • @jsg said:

    @dev_vps said:

    @jsg said:

    @dev_vps said:
    If I order a VPS from them, would you be willing to run those benchmark tests on my VPS to confirm that my VPS NVMe benchmark is similar to the one you tested with ?

    You seem to be a decent person, so: Yes, I would.

    Thank you so much, it is greatly appreciated.

    What configuration would you suggest for this test VPS? So that difference in benchmarks tests can not be attributed to the vps configuration

    I'd suggest the configuration that suits your needs. The processor is the same anyway and memory will make some difference but it won't turn a slow NVME into a fast one or vice versa. And after all I guess you want to know what kind of performance you can expect for your needs.

    You are welcome.

    I have just sent you PM with VPS details.
    Thanks, again, for your help with benchmark tests.
    Much appreciated

  • jsgjsg Member, Resident Benchmarker

    @dev_vps said:

    @jsg said:

    @dev_vps said:

    @jsg said:

    @dev_vps said:
    If I order a VPS from them, would you be willing to run those benchmark tests on my VPS to confirm that my VPS NVMe benchmark is similar to the one you tested with ?

    You seem to be a decent person, so: Yes, I would.

    Thank you so much, it is greatly appreciated.

    What configuration would you suggest for this test VPS? So that difference in benchmarks tests can not be attributed to the vps configuration

    I'd suggest the configuration that suits your needs. The processor is the same anyway and memory will make some difference but it won't turn a slow NVME into a fast one or vice versa. And after all I guess you want to know what kind of performance you can expect for your needs.

    You are welcome.

    I have just sent you PM with VPS details.
    Thanks, again, for your help with benchmark tests.
    Much appreciated

    No problem at all. You are welcome.

  • adlyadly Veteran
    edited August 2021

    @jsg said: OK, I've had with you assholes.

    Literally no-one asked for your benchmarks or opinions shitting on other benchmarks. Make the claims you do; expect them to be challenged.

  • Which disk I/O is better ?
    A or B

  • edited August 2021

    @dev_vps said:
    Which disk I/O is better ?
    A or B

    Depend on your usage. If you have a lot of small files then disk A is better.

  • cybertechcybertech Member
    edited August 2021

    what, using YABS fio to check I/O?

    hysterical-laughter.gif

  • @cybertech said:
    what, using YABS fio to check I/O?

    hysterical-laughter.gif

    Kindly let me know if there is a better script to benchmark I/O

  • adlyadly Veteran
    edited August 2021

    @jsg let me get this right - you're benchmarking on a rarely used server OS with a flawed understanding of the benchmark operation in Linux - and the benchmark script (YABS) is lacking?

    Beside, spouting technical BS and claiming everyone else are idiots is clearly a rational way to deal with criticism. Shows a real grip on the subject and its nuances.

  • AXYZEAXYZE Member
    edited August 2021

    @jsg said:

    @adly said:

    @jsg said:
    I clearly said that my test VMs were MXlinux with XFCE4.

    OK - I’ll check tomorrow. But not sure what relevance YABS not working on a rarely used desktop distro has to do with server/VPS benchmarking.

    (a) it was just a funny side note
    (b) It seems you'd be amazed to learn how many people do a graphical install
    (c) it was fio, not yabs I mentioned (in that side note)

    Oh, and "rarely used desktop distro"? Hahaha! Have a look at distrowatch. MXlinux is ranking number 1, debian is 7.

    You know distrowatch ranks distros based on page views> @dev_vps said:

    Which disk I/O is better ?
    A or B

    "A" for pretty everything. The limiting factor in both is 4k speeds. Sequentials are already awesome in A, in B theyre amazing, but you get these results on VPS because nobody needs that much speed.
    4k is a lot more important and thats why they limit it.

    Two people running Yabs would saturate sequentials of these drives in raid. Just two. But nobody will need this much speed so they just let it be, because besides benchmarks nobody will hit these speeds. 22MB/s in 4K can be saturated even in Minecraft server with mods and in databases that have a lot of relational queries. If you put Minecraft and databases into one server (hey, 4+ cores!) then good luck xD They will fight for IOPS and Contabo knows this. By limiting 4K IO nobody will use CPU that much so people wont notice that they are heavily oversold.

    Kinda funny - in minecraft forums people are complaining that 10 cores from Contabo isnt giving them lag-free experiemce. Then they switch to CHEAPER 2/3/4core Hetzner/BuyVM/Nexusbytes/netcup plan and problems go away :)

  • @AXYZE said:

    @dev_vps said:
    Which disk I/O is better ?
    A or B

    "A" for pretty everything. The limiting factor in both is 4k speeds. Sequentials are already awesome in A, in B theyre amazing, but you get these results on VPS because nobody needs that much speed.
    4k is a lot more important and thats why they limit it.

    Two people running Yabs would saturate sequentials of these drives in raid. Just two. But nobody will need this much speed so they just let it be, because besides benchmarks nobody will hit these speeds. 22MB/s in 4K can be saturated even in Minecraft server with mods and in databases that have a lot of relational queries. If you put Minecraft and databases into one server (hey, 4+ cores!) then good luck xD They will fight for IOPS and Contabo knows this. By limiting 4K IO nobody will use CPU that much so people wont notice that they are heavily oversold.

    Kinda funny - in minecraft forums people are complaining that 10 cores from Contabo isnt giving them lag-free experiemce. Then they switch to CHEAPER 2/3/4core Hetzner/BuyVM/Nexusbytes/netcup plan and problems go away :)

    I appreciate very much the details and feedback you have put in here. I have been windows guy for 20+ years and started using Linux only last year.

    A is RackNerd
    B is Contabo

  • @adly said:
    Well, I can say my penis is three times longer than any other penis measured. But I also don't have to show you my penis.

    Actually, it's six times longer but you're going to officially guesstimate three times.

    Thanked by 1adly
  • The disk space is quite a big hit...

  • jsgjsg Member, Resident Benchmarker
    edited August 2021

    @adly said:

    @jsg said: OK, I've had with you assholes.

    Literally no-one asked for your benchmarks or opinions shitting on other benchmarks. Make the claims you do; expect them to be challenged.

    (a) Literally wrong. Multiple providers asked for them and @jbiloh asked me to do benchmarks for/on LET/LEB.
    (b) Everyone is free to ignore benchmarks and reviews (s)he doesn't like.
    (c) Which part of "I've had it with you assholes" did you fail to grasp?

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @jsg said:
    (c) Which part of "I've had it with you assholes" did you fail to grasp?

    You’re a few months too early for the airing of grievances.

    Francisco

    Thanked by 2chocolateshirt 0xbkt
  • SagnikSSagnikS Member, Host Rep

    @jsg said: Oh, and "rarely used desktop distro"? Hahaha! Have a look at distrowatch. MXlinux is ranking number 1, debian is 7.

    Distrowatch is largely useless. It ranks distros based on the number of hits the Distrowatch page gets. The sole fact that MXlinux is #1 and Debian #7 is laughable.

Sign In or Register to comment.