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 3
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

13567

Comments

  • My lscpu is the same as posted above.

    Ive already submitted my cancellation with Contabo.

    What I expected from Contabo and really that quote

    "These NVMe are 3x faster than any other NVMe drives I've ever tested!"

    Should be changed to

    "These NVMe are 3x faster than any other NVMe drives I've ever tested! until Contabo restrict the service for paying customers"

  • jsgjsg Member, Resident Benchmarker

    @hennaboy said:
    My lscpu is the same as posted above.

    Ive already submitted my cancellation with Contabo.

    What I expected from Contabo and really that quote

    "These NVMe are 3x faster than any other NVMe drives I've ever tested!"

    Should be changed to

    "These NVMe are 3x faster than any other NVMe drives I've ever tested! until Contabo restrict the service for paying customers"

    Sorry, this is getting boring. I am expected to explain and defend everything yet the Contabo haters feel free to throw just about anything, incl. plenty BS.

    For a start: Using my default benchmark settings, the ones I always use (I like comparability) their NVMe were not 3 times faster but over 6 times faster.
    But that was on an (almost) empty node,so I lowered my guesstimate quite a lot to '3 times' plus I hammered the test node from 5 "background" VMs while running the benchmark on the test node.

    Then came a wave of, pardon me, worthless and meaningless yabs results. And again I made the effort to actually analyze it and to clearly lay out why and how it's worthless ... while some here just keeping throwing anything they happen to consider relevant at me.

    Shove it up your ... !

    I just yesterday benchmarked a test VM again, as it's highly likely fully loaded by now. The result? A bit better than what I had guesstimated.

  • adlyadly Veteran
    edited August 2021

    @jsg said: First a "funny" observation: fio couldn't be run on the small VM with 512 MB memory and even with 1 GB. I had to increase the memory to 1.5 GB in order to run fio on the "small VM because it seems that fio needs at least 512 MB free memory to run at all with the yabs parameters.

    Not sure what you're doing, but a fresh minimal Debian install with 512MB RAM (no SWAP) and 1 vCore runs fio with the YABS parameters fine:

  • jsgjsg Member, Resident Benchmarker

    @adly said:

    @jsg said: First a "funny" observation: fio couldn't be run on the small VM with 512 MB memory and even with 1 GB. I had to increase the memory to 1.5 GB in order to run fio on the "small VM because it seems that fio needs at least 512 MB free memory to run at all with the yabs parameters.

    Not sure what you're doing, but a fresh minimal Debian install with 512MB RAM (no SWAP) and 1 vCore runs fio with the YABS parameters fine:

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

    Thanked by 1Arkas
  • adlyadly Veteran

    @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.

    Thanked by 1cybertech
  • jsgjsg Member, Resident Benchmarker

    @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.

  • ArkasArkas Moderator

    All my linux machines (PCs) run MX Linux

  • edited August 2021

    First time hearing mxlinux

    Edit: meh, it was based on debian, I am not interested to use desktop gui on linux..

  • HarmonyHarmony Member
    edited August 2021

    Another YABS

    Processor  : AMD EPYC 7282 16-Core Processor
    CPU cores  : 4 @ 2794.748 MHz
    AES-NI     : ✔ Enabled
    VM-x/AMD-V : ❌ Disabled
    RAM        : 29.4 GiB
    Swap       : 2.0 GiB
    Disk       : 195.8 GiB
    
    fio Disk Speed Tests (Mixed R/W 50/50):
    ---------------------------------
    Block Size | 4k            (IOPS) | 64k           (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 23.30 MB/s    (5.8k) | 278.66 MB/s   (4.3k)
    Write      | 23.30 MB/s    (5.8k) | 280.13 MB/s   (4.3k)
    Total      | 46.61 MB/s   (11.6k) | 558.79 MB/s   (8.7k)
               |                      |
    Block Size | 512k          (IOPS) | 1m            (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 917.32 MB/s   (1.7k) | 1.03 GB/s     (1.0k)
    Write      | 966.06 MB/s   (1.8k) | 1.10 GB/s     (1.0k)
    Total      | 1.88 GB/s     (3.6k) | 2.13 GB/s     (2.0k)
    

    Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm rep_good nopl cpuid extd_apicid tsc_known_freq pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw perfctr_core ssbd ibrs ibpb stibp vmmcall fsgsbase tsc_adjust bmi1 avx2 smep bmi2 rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves clzero xsaveerptr wbnoinvd arat umip rdpid arch_capabilities

    @jsg what tests should I run? I will happily post them here.

    Also someone else was running benchmark tests for over 48 hours and their CPU power got cut in half, which is to be expected if you're not ordering a VDS.

    sysbench --test=cpu --num-threads=8 --cpu-max-prime=50000 run

    Before:

    CPU speed:
        events per second:  1305.07
    
    General statistics:
        total time:                          10.0059s
        total number of events:              13060
    
    Latency (ms):
             min:                                    5.93
             avg:                                    6.13
             max:                                   26.92
             95th percentile:                        6.21
             sum:                                79995.85
    

    After:
    CPU speed:
    events per second: 678.52

    General statistics:
        total time:                          10.0518s
        total number of events:              6821
    
    Latency (ms):
             min:                                    5.95
             avg:                                   11.72
             max:                                   88.22
             95th percentile:                       51.94
             sum:                                79975.38
    
    Threads fairness:
        events (avg/stddev):           852.6250/41.47
        execution time (avg/stddev):   9.9969/0.02
    
  • @jsg said:

    @hennaboy said:
    My lscpu is the same as posted above.

    Ive already submitted my cancellation with Contabo.

    What I expected from Contabo and really that quote

    "These NVMe are 3x faster than any other NVMe drives I've ever tested!"

    Should be changed to

    "These NVMe are 3x faster than any other NVMe drives I've ever tested! until Contabo restrict the service for paying customers"

    Sorry, this is getting boring. I am expected to explain and defend everything yet the Contabo haters feel free to throw just about anything, incl. plenty BS.

    For a start: Using my default benchmark settings, the ones I always use (I like comparability) their NVMe were not 3 times faster but over 6 times faster.
    But that was on an (almost) empty node,so I lowered my guesstimate quite a lot to '3 times' plus I hammered the test node from 5 "background" VMs while running the benchmark on the test node.

    Then came a wave of, pardon me, worthless and meaningless yabs results. And again I made the effort to actually analyze it and to clearly lay out why and how it's worthless ... while some here just keeping throwing anything they happen to consider relevant at me.

    Shove it up your ... !

    I just yesterday benchmarked a test VM again, as it's highly likely fully loaded by now. The result? A bit better than what I had guesstimated.

    Shove it up my?

    I would be embarrassed to be quoted by Contabo on this. Credibility taking it up the backside which is where you can shove it.

    I should take this to Paypal and dispute it for being missold. Absolute rubbish but of course everyone else's tests and scripts must be worthless. Only yours.

    Absolute crock of sh*t you getting % from Contabo?

  • adlyadly Veteran

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

    MX Linux contains an outdated version of fio in the repos, it works fine if you use the fio version supplied with YABS (3.27):

    I'm not sure how YABS (or fio) can be to blame for issues if you're using very old versions.

    @jsg said:
    (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.

    (a) It shows flawed thinking to blame the YABS fio parameters when it's more likely the machine configuration/OS specific, which puts the rest of your claims in doubt.
    (b) Perhaps they do, but not on a 512MB VPS, surely?
    (c) I'm aware it was fio, but you blamed the YABS parameters in the side note.

    I wasn't aware that MX Linux is so popular on the desktop, but unless this is popular with server/VPS providers then it isn't really relevant for benchmarking.

  • adlyadly Veteran

    @jsg said: I can't because FreeBSD doesn't really support fio.

    Just for extra fun - latest fio installed from FreeBSD ports running just fine using the posixaio engine:

    Thanked by 1chocolateshirt
  • adlyadly Veteran
    edited August 2021

    Just some more afternoon fun...

    @jsg said: uses 64 (!) "IO units" that is, is does 64 reads and writes at the same time which normal applications almost never do; to simulate database servers though it may be useful. [Mine does (series of) single reads or writes].

    The I/O depth specifies how many I/Os fio will submit to the OS before waiting for earlier I/O to be acknowledged - that is, in this case there can be 64 unacknowledged I/Os before fio will wait before submitting any further I/Os. It is no guarantee that 64 concurrent I/Os are actually occurring, some could actually be queued by the kernel.

    You also missed that YABS is specifying 2 jobs, so it's actually submitting 128 I/Os to the OS (64 per job/process). Perhaps this is not normal, and is more of a test suited to database servers, but then the argument should surely be to reduce I/O depth and increase job number (more processes, less I/O per process) to be more realistic?

    @jsg said: uses 4 threads ("jobs" in their lingo) which only disk IO heavy applications would do but most applications would not. And that, well noted, in addition to using 64 "IO units"! [My disk testing runs single threaded as most applications do it] .

    Ignoring everything else discussed here, unless --threads is passed to fio it forks a new process, it does not create threads by default. Also YABS specifies 2 jobs, not 4.

    @jsg said: heftily reduces [get]timeofday() calls as well as result precision. That is absolutely not what one wants in a benchmark [Mine does all test with microsecond granularity].

    fio highlights that a large amount of gettimeofday() calls can impact performance at high IOPS rates.

  • dedicatserver_rodedicatserver_ro Member, Host Rep

    @jsg , Server Review King said: I'm not saying that my benchmark is the only true and good one, but it's something that delivers what all the scripts I know of don't deliver.

    • false results ;)
    • wrong expectations 👍
    • wrong storage classifications 👏
  • jsgjsg Member, Resident Benchmarker

    @adly

    To cut it short: you are on a mission to somehow prove me and my benchmark wrong, no matter how, no matter the means. For example, you blew up a funny side note totally out of proportion because it suited your mission. At the same time you conveniently "forgot" to do tests that don't suit you.

    And you got your reward. The ever same jsg haters who click "Thanks" whenever and no matter about what someone contradicts me. Congrats.

    Unfortunately though, that's not my style and I don't do dirty pit fights.

    If you are so great and know so much why don't you just write a better benchmark program than I did?

    @dedicatserver_ro

    Thanks for taking some time in between spying on customers ;)

    Thanked by 1Arkas
  • adlyadly Veteran
    edited August 2021

    @jsg said:
    @adly

    To cut it short: you are on a mission to somehow prove me and my benchmark wrong, no matter how, no matter the means. For example, you blew up a funny side note totally out of proportion because it suited your mission. At the same time you conveniently "forgot" to do tests that don't suit you.

    And you got your reward. The ever same jsg haters who click "Thanks" whenever and no matter about what someone contradicts me. Congrats.

    Unfortunately though, that's not my style and I don't do dirty pit fights.

    If you are so great and know so much why don't you just write a better benchmark program than I did?

    I'm not on a mission to prove you or you're benchmark wrong (I've never mentioned or even looked at your benchmarking tool once). If I have forgot any tests, it's not intentional - have you tried following one of your posts? I am more than happy to fill in any gaps.

    As for writing my own program, I don't think it's necessary to write a better benchmark program, the existing tools can be improved and are mostly OK.

    My goal is purely to call out BS where I see it, and I'm sad to say that you're appearing to live up to the W.C. Fields quote: “If you can't dazzle them with brilliance, baffle them with bullshit.” However, I'd love to be proven wrong.

  • dev_vpsdev_vps Member
    edited August 2021

    just one question
    @jsg @adly and others -
    is Contabo throttling I/O limit?

  • adlyadly Veteran

    @dev_vps I've no idea - never used them and don't feel qualified to comment on that. There are a number of independent benchmarks in this thread that suggest as much, but make your own opinion.

  • jsgjsg Member, Resident Benchmarker

    @adly said:
    My goal is purely to call out BS where I see it, and I'm sad to say that you're appearing to live up to the W.C. Fields quote: “If you can't dazzle them with brilliance, baffle them with bullshit.” However, I'd love to be proven wrong.

    Nice try. One of your major targets was a, I quote, "funny side note".

  • @contabo_m
    Would you please kindly confirm if there is any I/O limit set for new VPS that uses NVMe based storage?

    If there is a limit, what is that limit?

    Thanks so much

  • adlyadly Veteran

    @jsg said: Nice try. One of your major targets was a, I quote, "funny side note".

    Let me clarify, it's not a target, it's proof that the claims you made in the "funny side note" are wrong. Nothing more, nothing less. I have no axe to grind; I saw a false statement, and I responded to it.

  • jsgjsg Member, Resident Benchmarker
    edited August 2021

    @dev_vps said:
    just one question
    @jsg @adly and others -
    is Contabo throttling I/O limit?

    At the end of the day, only Contabo can answer that. Btw. some throttling/limiting is OK, even required and immanent with VPS (and even VDS). The question is (a) whether Contabo (as alleged by some) really did provide me with a "special" VPS for testing (I'm doubting it but I'm following those allegations adf try hard to verify), and (b) whether the throttling/limiting they do is unreasonably harsh.

  • @jsg said:

    @dev_vps said:
    just one question
    @jsg @adly and others -
    is Contabo throttling I/O limit?

    At the end of the day, only Contabo can answer that. Btw. some throttling/limiting is OK, even required and immanent with VPS (and even VDS). The question is (a) whether Contabo (as alleged by some) really did provide me with a "special" VPS for testing (I'm doubting it but I'm following those allegations adf try hard to verify), and (b) whether the throttling/limiting they do is unreasonably harsh.

    Thanks and that is why I have posted the question directly to @contabo_m

    Thanked by 1adly
  • jsgjsg Member, Resident Benchmarker

    UPDATE

    I got feedback from Contabo.

    He (my contact) said that to the best of his knowledge my test VPSs were not in any way special. He also told me that Contabo does some throttling/limiting with "noisy" VPS/customers that is, with VPSs that often use (considerably) more than their fair share.
    That response matches well with what is common practice in the industry.

    There is a grain of salt though: my contact is pretty high up at Contabo and what he says is based on what the techies tell him. Based on multiple rather technical information I got at multiple occasions (after asking rather detailed tech. questions) though it seems that the internal communication lines work fine. The answers I got evidently came from tech people (and my contact just relayed them to me) and were of good quality.

    As this is an important matter (for us at LET) I'll try to follow up a bit more.

    Relevant side note: It's well known from diverse providers that the customers who are limited due to their (basically ab)use of the system tend to make lots of noise in forums and what they allege should be checked.

  • dev_vpsdev_vps Member
    edited August 2021

    @jsg said:
    UPDATE

    I got feedback from Contabo.

    He (my contact) said that to the best of his knowledge my test VPSs were not in any way special. He also told me that Contabo does some throttling/limiting with "noisy" VPS/customers that is, with VPSs that often use (considerably) more than their fair share.
    That response matches well with what is common practice in the industry.

    There is a grain of salt though: my contact is pretty high up at Contabo and what he says is based on what the techies tell him. Based on multiple rather technical information I got at multiple occasions (after asking rather detailed tech. questions) though it seems that the internal communication lines work fine. The answers I got evidently came from tech people (and my contact just relayed them to me) and were of good quality.

    As this is an important matter (for us at LET) I'll try to follow up a bit more.

    Relevant side note: It's well known from diverse providers that the customers who are limited due to their (basically ab)use of the system tend to make lots of noise in forums and what they allege should be checked.

    question is -

    is there any global I/O limit that applies to ALL the customers?
    Again, to all the customers

    I like numbers , .... not generic response - fair usage, common industry practice etc

    for example -
    1. using average cpu resources around 30% (measured over 6 hours) -- is it fair usage or over the limit
    2. what is the average I/O limit (measured over 6 hours) that is considered over the limit of fair usage?
    3. What are the key parameters that are used to determine a customer is “noisy”?

    cc - @contabo_m

  • jsgjsg Member, Resident Benchmarker

    @dev_vps said:
    question is -

    is there any global I/O limit that applies to ALL the customers?
    Again, to all the customers

    That's something I also try to find out, but ...

    I like numbers , .... not generic response - fair usage, common industry practice etc

    for example -
    1. using average cpu resources around 30% (measured over 6 hours) -- is it fair usage or over the limit

    Maybe we'll get an answer - but probably not. That question is one that most providers don't like to answer at all.

    1. What are the key parameters that are used to determine a customer is “noisy”?

    I happen to know that they do not seem to "quickly fire from the hip" nor do they bureaucratically hit the "limit" button the second one is over the allowed rate no matter how little. I guess the answer is "they limit when a customer/VPS repeatedly crosses the limit significantly".

  • dev_vpsdev_vps Member
    edited August 2021

    @jsg said:

    @dev_vps said:
    question is -

    is there any global I/O limit that applies to ALL the customers?
    Again, to all the customers

    That's something I also try to find out, but ...

    I like numbers , .... not generic response - fair usage, common industry practice etc

    for example -
    1. using average cpu resources around 30% (measured over 6 hours) -- is it fair usage or over the limit

    Maybe we'll get an answer - but probably not. That question is one that most providers don't like to answer at all.

    I ask the provider, directly, before placing my order, about the allowed cpu % usage (over 6-8 hours rolling window) and order is placed only AFTER they agree with my requested cpu % limit

    in short, if the provider is NOT ready to share such details that will make the customer comfortable, I would avoid that provider. But that is just me.

    cc - @contabo_m

  • @jsg said: @adly
    To cut it short: you are on a mission to somehow prove me and my benchmark wrong

    Nobody is on a mission to prove you wrong, because you are wrong, just like that time when you were insisting @cociu is not dead and wasn't lying about the involucration saga.

  • jsgjsg Member, Resident Benchmarker
    edited August 2021

    @stevewatson301 said:

    @jsg said: @adly
    To cut it short: you are on a mission to somehow prove me and my benchmark wrong

    Nobody is on a mission to prove you wrong, because you are wrong, just like that time when you were insisting @cociu is not dead and wasn't lying about the involucration saga.

    Uhum, if you say so ...

    Funny logic btw. "You are wrong, not even an attempt of proving needed". Oh well, ...
    I considered to respond to a post of @deank (iirc) who said (in another thread) that the average IQ on LET is 7 and asking him to increase that at least to 70 (with an IQ of 83 even the army considers one basically useless), but I guess I was right to not post that. I still think better of LET but it seems I've been too optimistic.

    And you, sir, have 3 arms and 5 legs. I'm not even trying to prove that because it is so because someone said so. Not a doctor, not even someone with a major in bio, but we aren't picky around here.

  • dev_vpsdev_vps Member
    edited August 2021

    @jsg @adly @stevewatson301

    is there any linux script that can be used to determine if the VPS has any throttled IO limit set ?

Sign In or Register to comment.