Howdy, Stranger!

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


Contabo is giving it's SSD storage by double - 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 is giving it's SSD storage by double

2»

Comments

  • @AXYZE said:

    @add_iT said:

    @Astro said:

    @add_iT said:

    @Astro said:

    @add_iT said:
    But it is 200Mbit/s :/

    More than adequate for most tasks. I stream 4k without encoding over Plex on the sys 250 mbps

    Yes

    That is the only deal breaker for me

    What's your use case?

    YABS 24/7 B)

    Btw any similiar offer like contabo with same price but with 1Gbit/s minimal ?

    Location does't matter

    Looking something like 8GB ram <= 60 Euro /y

    netcup https://www.netcup.eu/vserver/vps.php
    I can give you discount code too, just hit me up on PM.

    I am curious what is total price for one year for VPS 1000 G9 after discount applied

  • DediRockDediRock Member, Patron Provider

    Pretty much standard practice in Germany.

    Thanked by 1jason5545
  • is contabo vps even raid 10

  • jsgjsg Member, Resident Benchmarker
    edited February 2022

    @Astro said:

    @add_iT said:
    But it is 200Mbit/s :/

    More than adequate for most tasks. I stream 4k without encoding over Plex on the sys 250 mbps

    Double Yes. Connectivity seems indeed to be limited to 200 Mb/s (at least in DE, which is what I just checked), but yes again, 200 Mb/s is plenty enough for most use cases.

    As for the NVMes they are not at all slow, quite the contrary, but they have a high spread. For example on my (DE) NVMe I see 4k4T IOPS (occasionally) going as low as about 4000, occasionally as high as 8000+ and average seems to be about 6000. I happen to know the exact model and it's among the top-performers and I know that Contabo does not brutally oversell but alas they seem to have problems with their automatic monitoring and resource management (aka known as reigning in on abusers).

    Btw, on my us-american Contabo VPS (with SSD) they indeed seem to have disk-IO limiting at work (500 IOPS), so the numbers are nicely stable (very low spread) but quite low.

    But still, 4 Epyc vCores, 8 GB memory, and 200 GB SSD for about €5 per month is a very hard to beat offer.

  • @jsg said:

    @Astro said:

    @add_iT said:
    But it is 200Mbit/s :/

    More than adequate for most tasks. I stream 4k without encoding over Plex on the sys 250 mbps

    Double Yes. Connectivity seems indeed to be limited to 200 Mb/s (at least in DE, which is what I just checked), but yes again, 200 Mb/s is plenty enough for most use cases.

    As for the NVMes they are not at all slow, quite the contrary, but they have a high spread. For example on my (DE) NVMe I see 4k4T IOPS (occasionally) going as low as about 4000, occasionally as high as 8000+ and average seems to be about 6000. I happen to know the exact model and it's among the top-performers and I know that Contabo does not brutally oversell but alas they seem to have problems with their automatic monitoring and resource management (aka known as reigning in on abusers).

    Btw, on my us-american Contabo VPS (with SSD) they indeed seem to have disk-IO limiting at work (500 IOPS), so the numbers are nicely stable (very low spread) but quite low.

    But still, 4 Epyc vCores, 8 GB memory, and 200 GB SSD for about €5 per month is a very hard to beat offer.

    Indeed a great offer but I am getting a GB5 multi core score of 1100 which seems pretty low. Reached out to them. Lets see what they say.

  • jsgjsg Member, Resident Benchmarker

    @Astro said:
    Indeed a great offer but I am getting a GB5 multi core score of 1100 which seems pretty low. Reached out to them. Lets see what they say.

    Based on my own benchmarks (I already did multiple runs) I'd say Contabo's Epyc vCores perform roughly like decent (but not great) E5-26xx v4 vCores - with one potentially major difference: considerably faster OpenSSL RSA and AES.

    I guess it depends on one's perspective. Some are disappointed ("Meee, Ryzen 5000 is much faster!") and some, incl. myself, are OK or even happy with the performance. I guess it's a bit like '200 Mb/s connectivity' thing; some react as if 200 Mb/s were an insult while others, incl. myself, think it's really good enough for 90+% of use cases.

    Don't get me wrong, I really like Ryzen 5000 (or even "only" 3000) based systems, but frankly, this is LET and if someone really needs more than Epyc or E5-2600 v4 performance and more than 200 Mb/s connectivity he's probably in the wrong community anyway.
    I happen to co-run a relatively busy and prominent community - on a (dual) "stone-age" 2450L dedi. And we're not even using half its cores.

  • @jsg said:

    @Astro said:
    Indeed a great offer but I am getting a GB5 multi core score of 1100 which seems pretty low. Reached out to them. Lets see what they say.

    Based on my own benchmarks (I already did multiple runs) I'd say Contabo's Epyc vCores perform roughly like decent (but not great) E5-26xx v4 vCores - with one potentially major difference: considerably faster OpenSSL RSA and AES.

    I guess it depends on one's perspective. Some are disappointed ("Meee, Ryzen 5000 is much faster!") and some, incl. myself, are OK or even happy with the performance. I guess it's a bit like '200 Mb/s connectivity' thing; some react as if 200 Mb/s were an insult while others, incl. myself, think it's really good enough for 90+% of use cases.

    Don't get me wrong, I really like Ryzen 5000 (or even "only" 3000) based systems, but frankly, this is LET and if someone really needs more than Epyc or E5-2600 v4 performance and more than 200 Mb/s connectivity he's probably in the wrong community anyway.
    I happen to co-run a relatively busy and prominent community - on a (dual) "stone-age" 2450L dedi. And we're not even using half its cores.

    It is not that I'm satisfied, look here https://talk.lowendspirit.com/discussion/3770/contabo-400gb-free-ssd-if-you-order-today

    Huge disparity in CPU performance.

  • jsgjsg Member, Resident Benchmarker
    edited February 2022

    @Astro said:

    @jsg said:

    @Astro said:
    Indeed a great offer but I am getting a GB5 multi core score of 1100 which seems pretty low. Reached out to them. Lets see what they say.

    Based on my own benchmarks (I already did multiple runs) I'd say Contabo's Epyc vCores perform roughly like decent (but not great) E5-26xx v4 vCores - with one potentially major difference: considerably faster OpenSSL RSA and AES.

    I guess it depends on one's perspective. Some are disappointed ("Meee, Ryzen 5000 is much faster!") and some, incl. myself, are OK or even happy with the performance. I guess it's a bit like '200 Mb/s connectivity' thing; some react as if 200 Mb/s were an insult while others, incl. myself, think it's really good enough for 90+% of use cases.

    Don't get me wrong, I really like Ryzen 5000 (or even "only" 3000) based systems, but frankly, this is LET and if someone really needs more than Epyc or E5-2600 v4 performance and more than 200 Mb/s connectivity he's probably in the wrong community anyway.
    I happen to co-run a relatively busy and prominent community - on a (dual) "stone-age" 2450L dedi. And we're not even using half its cores.

    It is not that I'm satisfied, look here https://talk.lowendspirit.com/discussion/3770/contabo-400gb-free-ssd-if-you-order-today

    Huge disparity in CPU performance.

    Well if you care about those opinions then stay away.

    I don't, I don't give a flying f_ck on much of those numbers and opinions. Let me tell you why:

    For a start, again,we're talking about a VPS with 4 vCores and not crappy one from 2015 but relatively current ones in relatively new machines, plenty memory, generous disk space, at an almost ridiculously low sales price. Unlike most LES "experts" and "benchmark kings" I happen to know those machines, I happen to know what NVMes and SSDs are in there, what memory sticks are in there, how many VPS are on a node (hint: no, they are not brutally overselling) and in summary I know that Contabo weren't cheap shots when they got their hardware. Contabo provided me with quite generous access and insight; about the only thing they kept hidden - and rightly so - was customer related info/data.

    Next point: what are those "experts" blathering about? Are we playing a number-porn game? If so, I'm out, I'm not at all interested in number porn. My benchmark program was designed and written with a clear purpose - and that was not number-porn; it was to get real world numbers with the aim of being able to judge what I/we can expect from a given VPS.

    "Steal factor"? I don't give a flying f_ck. My benchmarking get real world "net" performance. So, if a VPS has a high steal then my benchmark shows lower processor performance just like e.g. nginx or mysql or whatever would.

    What I see on all my Contabo VPS is what I call "decent 26xx v4 numbers" but with higher crypto performance (which is normal and expected from an Epyc). And I see that the results for all 4 vCores is about or quite close to 3 times the single-core performance, which is a very decent result (and also Zen typical).

    Does Contabo have a problem? Yes, you bet. But that problem isn't brutally overselling (as often alleged) or crappy hardware. That problem is an unlucky combination of abusers and Contabo (IMO) having a monitoring/management problem.

    The end result and summary is that you indeed don't get the full performance expected from an Epyc, but you still get very decent performance and a lot of bang per buck.

    Would I recommend a Contabo VPS for critical business use? Nope. For really high load? Nope. For some private project or a local restaurants or bicycle shops site? Yep, definitely.

  • @jsg said: My benchmark program was designed and written with a clear purpose - and that was not number-porn; it was to get real world numbers with the aim of being able to judge what I/we can expect from a given VPS.

    Any chance you could share it?

  • jsgjsg Member, Resident Benchmarker

    @Astro said:

    @jsg said: My benchmark program was designed and written with a clear purpose - and that was not number-porn; it was to get real world numbers with the aim of being able to judge what I/we can expect from a given VPS.

    Any chance you could share it?

    I once did, I even open-sourced it. What I got for it was e.g. complaints about not being on github. Just ask the many "experts" for the benchmarking software they created and published on github, or no, wait, they didn't.

  • @jsg said:

    @Astro said:

    @jsg said: My benchmark program was designed and written with a clear purpose - and that was not number-porn; it was to get real world numbers with the aim of being able to judge what I/we can expect from a given VPS.

    Any chance you could share it?

    I once did, I even open-sourced it. What I got for it was e.g. complaints about not being on github. Just ask the many "experts" for the benchmarking software they created and published on github, or no, wait, they didn't.

    Mason did ;)

    I don't need GitHub! You can DM if you'd like. Thank you!

    Thanked by 1Goodtesteronline
  • @jsg can you confirm Contabo VPS use Raid 10? They have to cut something with that price.

  • jsgjsg Member, Resident Benchmarker

    @Goodtesteronline said:
    @jsg can you confirm Contabo VPS use Raid 10? They have to cut something with that price.

    No, unfortunately I can't, which isn't Contabo's fault but my having forgotten. I seem to remember though that they use some form of Raid but whether it's 10 I sadly don't remember.

  • mkvmkv Member
    edited February 2022

    I purchased one from them (the 4-core 2x200 GB SSD version, Rocky Linux). The installation was done in 12 hours instead of the three hours promised.

    From the very start I noticed some hiccups but was not worried. Soon I noticed a considerable sluggishness and some dropouts due to timeouts. Finally, I ran the SSD benchmark test, which resulted in 20 MB/s. That looked bad.

    Dnf updates started extremely slowly and finally I decided to measure the execution times. A simple listing (dnf list installed) that normally takes a second or two took more than two minutes. A simple dnf command without parameters:

    real 1m1.631s
    user 0m0.734s
    sys 0m0.634s

    All executable files start very slowly, here you can see another "nice" example:

    time certbot --version
    certbot 1.22.0

    real 0m53.495s
    user 0m2.345s
    sys 0m6.793s

    There is no ticketing system, just phone and email. No real person has so far answered my ticket. Have a look at their Facebook feed; I am not alone with my problems.

  • @mkv said:
    I purchased one from them (the 4-core 2x200 GB SSD version, Rocky Linux). The installation was done in 12 hours instead of the three hours promised.

    From the very start I noticed some hiccups but was not worried. Soon I noticed a considerable sluggishness and some dropouts due to timeouts. Finally, I ran the SSD benchmark test, which resulted in 20 MB/s. That looked bad.

    Dnf updates started extremely slowly and finally I decided to measure the execution times. A simple listing (dnf list installed) that normally takes a second or two took more than two minutes. A simple dnf command without parameters:

    real 1m1.631s
    user 0m0.734s
    sys 0m0.634s

    All executable files start very slowly, here you can see another "nice" example:

    time certbot --version
    certbot 1.22.0

    real 0m53.495s
    user 0m2.345s
    sys 0m6.793s

    There is no ticketing system, just phone and email. No real person has so far answered my ticket. Have a look at their Facebook feed; I am not alone with my problems.

    I got a reply via the email in 29 hrs. Its the weekend so it might be a while.

  • Sigh. I gave them my customer ID and they cannot locate me in the system because "we cannot find your email". Yes, I used another email account which is why I included my customer ID... but, yes, this is their "ticket system" at its very best.

  • As a proof, I sent the support measurements of execution times, some of which were 100x **longer than in a low-end VPS. Their answer was that they **did not understand what I was writing. While I was writing my last message in easy-reading format targeted for children, the VPS suddenly went unresponsive and had to be restarted. Then, after a few minutes, it was unresponsive again.

    Between those ghost boots/lock-ups/whatever, the server worked as fast as it should have done from hour one.

    The VPS panel shows that it is running, but it is not responsive and its status is "unknown". This is a hide-and-seek game, at best.

    Tomorrow will probably reveal more about the saga. This is very interesting.

    Thanked by 2darkimmortal pbx
  • I installed KeyHelp on it and it was installed with KeyHelp Pro 🤔

  • @jcarlo9 said:
    I installed KeyHelp on it and it was installed with KeyHelp Pro 🤔

    Maybe they partnered up, or it's a free trial that automatically gets activated? Would be cool if it was included :)

  • mkvmkv Member
    edited March 2022

    Well, it took them two weeks to get their backend stable enough (no more lock-ups and reboots). However, the "SSD" speed is less than impressive:

    [root@vmiXXXXX]# dd if=/dev/zero of=/tmp/test1.img bs=1G count=1 oflag=dsync
    1+0 records in
    1+0 records out
    1073741824 bytes (1.1 GB, 1.0 GiB) copied, 40.9938 s, 26.2 MB/s

    The speed fluctuates from "nearly unusable" to "pretty good". Sometimes disk access involves huge delays, ending up in messages I have never seen before:

    [root@vmiXXXXX]# adduser johndoe
    The memcache was not invalidated by NSS responder.

    The reason behind this is

    "If sssd is busy or the system overall is overloaded, it can take much longer to clear the cache. Typically, this happens if you are adding or deleting users in large batches. "

    In the Contabo case, it takes place when adding just one user, which may take seconds.

    Oh, well. This is DEFINITELY NOT a production server, but it may have its place as a backup one.

  • @mkv said:
    Well, it took them two weeks to get the system stable enough (no more lock-ups and reboots). However, the "SSD" speed is less than impressive:

    [root@vmiXXXXX]# dd if=/dev/zero of=/tmp/test1.img bs=1G count=1 oflag=dsync
    1+0 records in
    1+0 records out
    1073741824 bytes (1.1 GB, 1.0 GiB) copied, 40.9938 s, 26.2 MB/s

    The speed fluctuates from "nearly unusable" to "pretty good". Sometimes disk access involves huge delays, ending up in messages I have never seen before:

    [root@vmiXXXXX]# adduser johndoe
    The memcache was not invalidated by NSS responder.

    The reason behind this is

    "If sssd is busy or the system overall is overloaded, it can take much longer to clear the cache. Typically, this happens if you are adding or deleting users in large batches. "

    In the Contabo case, it takes place when adding just one user, which may take seconds.

    Oh, well. This is DEFINITELY NOT a production server, but it may have its place as a backup one.

    this is sequential speed, 26MB/s is worse than HDD. even scammy pacificrack was never this bad.

    i wouldn't even use it for backup. Have you seen how the storage vps from better LE providers here perform?

    could you try the same command while having another window with top running and screenshot it?

    can you also post a YABS?

  • @mkv said:
    Well, it took them two weeks to get their backend stable enough (no more lock-ups and reboots). However, the "SSD" speed is less than impressive:

    [root@vmiXXXXX]# dd if=/dev/zero of=/tmp/test1.img bs=1G count=1 oflag=dsync
    1+0 records in
    1+0 records out
    1073741824 bytes (1.1 GB, 1.0 GiB) copied, 40.9938 s, 26.2 MB/s

    The speed fluctuates from "nearly unusable" to "pretty good". Sometimes disk access involves huge delays, ending up in messages I have never seen before:

    [root@vmiXXXXX]# adduser johndoe
    The memcache was not invalidated by NSS responder.

    The reason behind this is

    "If sssd is busy or the system overall is overloaded, it can take much longer to clear the cache. Typically, this happens if you are adding or deleting users in large batches. "

    In the Contabo case, it takes place when adding just one user, which may take seconds.

    Oh, well. This is DEFINITELY NOT a production server, but it may have its place as a backup one.

    Which location did you choose?

  • mkvmkv Member
    edited March 2022

    The speed test results are below. I am fully aware of what the speed should be, having used a couple of RamNode HDD ones in the past. The numbers below are not even telling the truth. It is alarming that whenever you access a file in the file system that cannot be found in the OS disk cache, it may take up to several seconds to access it. The second time, the access speed is pretty much OK - until it vanishes from the disk cache.

    Overall, the "SSD" operation feels as if the data was first fetched from a storage to some kind of intermediate cache, a bit like getting it from an old-fashioned tape.

    There is no way I could use this as a mirror server for the online shops we are running. Moreover, after the first two weeks I am more than suspicious. I will probably use it as a secondary backup server for data and emails.

    Today, I was astonished to see that Contabo is offering a new service, an affordable object cache, a bit like the Amazon one. Call me paranoid, but I fear that this "400 GB SSD" has something to do with the "object storage", with its horribly bad performance. Whatever their "SSD space" is, it is **definitely ** not a normal SSD solution.

    fio Disk Speed Tests (Mixed R/W 50/50):
    ---------------------------------
    Block Size | 4k            (IOPS) | 64k           (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 5.30 MB/s     (1.3k) | 67.51 MB/s    (1.0k)
    Write      | 5.32 MB/s     (1.3k) | 67.82 MB/s    (1.0k)
    Total      | 10.62 MB/s    (2.6k) | 135.33 MB/s   (2.1k)
               |                      |
    Block Size | 512k          (IOPS) | 1m            (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 127.55 MB/s    (249) | 98.85 MB/s      (96)
    Write      | 134.32 MB/s    (262) | 105.43 MB/s    (102)
    Total      | 261.88 MB/s    (511) | 204.29 MB/s    (198)
    
  • @mkv said: I fear that this "400 GB SSD" has something to do with the "object storage", with its horribly bad performance.

    To be fair their object storage is also crap. It's just a basic Swift setup running behind Kong and Cloudflare, and that ratelimits GETs at 1 request/s :lol:

  • @mkv said:
    The speed test results are below. I am fully aware of what the speed should be, having used a couple of RamNode HDD ones in the past. The numbers below are not even telling the truth. It is alarming that whenever you access a file in the file system that cannot be found in the OS disk cache, it may take up to several seconds to access it. The second time, the access speed is pretty much OK - until it vanishes from the disk cache.

    Overall, the "SSD" operation feels as if the data was first fetched from a storage to some kind of intermediate cache, a bit like getting it from an old-fashioned tape.

    There is no way I could use this as a mirror server for the online shops we are running. Moreover, after the first two weeks I am more than suspicious. I will probably use it as a secondary backup server for data and emails.

    Today, I was astonished to see that Contabo is offering a new service, an affordable object cache, a bit like the Amazon one. Call me paranoid, but I fear that this "400 GB SSD" has something to do with the "object storage", with its horribly bad performance. Whatever their "SSD space" is, it is **definitely ** not a normal SSD solution.

    fio Disk Speed Tests (Mixed R/W 50/50):
    ---------------------------------
    Block Size | 4k            (IOPS) | 64k           (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 5.30 MB/s     (1.3k) | 67.51 MB/s    (1.0k)
    Write      | 5.32 MB/s     (1.3k) | 67.82 MB/s    (1.0k)
    Total      | 10.62 MB/s    (2.6k) | 135.33 MB/s   (2.1k)
               |                      |
    Block Size | 512k          (IOPS) | 1m            (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 127.55 MB/s    (249) | 98.85 MB/s      (96)
    Write      | 134.32 MB/s    (262) | 105.43 MB/s    (102)
    Total      | 261.88 MB/s    (511) | 204.29 MB/s    (198)
    

    i would look at the IOWait then. i have a storage vps right now with 90+ IOwait it takes forever to run commands

  • mkvmkv Member
    edited March 2022

    Well, IOWait and its friends go beyond my comfort zone. I am, after all, a basic server administrator (more of a software developer). I have installed at least two dozen virtual servers so far, and nothing has been as bad as this one. Wherever the culprit is hiding, I have no interest in spending hours searching for it. What I have, after all, is a VPS and the hardware is hidden behind the shades. One can only make guesses, at best.

    I cannot do much about the OS, either. I have tried two (Debian & Rocky), knowing both of them. It is up to the provider to make sure that the OS image provided works well enough. The lock-ups and frequent reboots by the hypervisor administrator made it clear that the backend is far from a working one.

    I will make one final request to the Contabo support and if nothing happens I will probably forget the next invoice. They have not been too helpful so far, after all. The only plus I can give them is for providing a badly performing but stable VPS instead of the unstable, badly performing one they had initially.

  • mkvmkv Member
    edited March 2022

    As of writing this, the system is dead again and cannot be started or restarted. The hypervisor is probably down as well because the VNC cannot be turned on. What works? Nothing. I have NEVER, not once, seen as bad a start as this one has been.

    UPDATE: "The network issue is caused by one of our Internet carriers. The fiber lines from their place to our data center are physically cut."

    I cannot believe this. Their service is bad and they have bad luck as well.

  • Is your vps located in Dusseldorf/DE ?
    there are problems with fibercut at datacenter

  • mkvmkv Member
    edited March 2022

    Yes, Düsseldorf. I noticed their FB message. And they have no backup connection, obviously. I have no idea how many data centres have backup ones, though. Moreover, they do not send any information to clients affected by ANY problems. I have seen already during my three weeks at Contabo. The hypervisor is down and no emails/FB posts. The fibre issue is the first one they have told us about, and that only in FB.

    From Contabo: "Server is only useful when it's reachable from the internet. To ensure that this is always the case we always install at least 2 independent fiber connections at each Data Center and we work with multiple carriers."

    I wonder where the second one is? Next to the cut one?

  • @mkv said:
    Yes, Düsseldorf. I noticed the message. And they have no backup connection, obviously. I have no idea how many data centres have backup ones, though.

    i think clouvider has

Sign In or Register to comment.