Howdy, Stranger!

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


Best affordable VDS providers: Hetnzer vs Netcup vs PHP Friends/Data Forest vs BMDS
New on LowEndTalk? Please Register and read our Community Rules.

Best affordable VDS providers: Hetnzer vs Netcup vs PHP Friends/Data Forest vs BMDS

Let's assume all VDS providers use the latest AMD EPYC cpus.

Hetzner mentioned EPYC 7003 series, very possibly 64 core 128 threads ones; Netcup mentioned EPYC 7702, 64 core 128 threads; php friends mentioned AMD EPYC 7452, 32 core 64 threads.

Hetzner's cheapest VDS, $20, 2 vcpu (thread); Netcup, $10, 4 vcpu (thread); php friends, $8, 2 vcpu.

Then each big server can host about 50-60 Hetnzer vds, 25-30 Netcup vds, 25-30 php friends vds.

Hetzner's network port is 10 gbps, Netcup 2.5 gbps, php friends 1 gbps.

Therefore, Hetzner's vds basic bandwidth is 10/50=200 mbps, Netcup 100 mbps, php friends 50 mbps.

If the big server is dual socket, then Hetzner's vds basic bandwidth is 10/100=100 mbps, Netcup 50 mbps, php friends 25 mbps.

In other words, Hetzner's cheapest vds has one 7763 cpu core, 4 gb ram, 100-200 mpbs 20 TB network, almost 100 mbps unmetered (32 tb). 7763 single thread passmark is 2600, one core 2 threads about 5000 passmark.

The 2022 Hetzner $20 vds is basically a $100 bmds in the year 2012. This is good enough for many website hosting except video streaming. But compared to Hetzner's bmds, its vds offer is not great.

Both Netcup and php friends need to increase their network port to 10 gbps.

«1

Comments

  • AXYZEAXYZE Member
    edited June 16

    @letlover said:
    Then each big server can host about 50-60 Hetnzer vds, 25-30 Netcup vds, 25-30 php friends vds.

    All of them host more instances per node.

    If the big server is dual socket, then Hetzner's vds basic bandwidth is 10/100=100 mbps, Netcup 50 mbps, php friends 25 mbps.

    No. You have 1Gbps DEDICATED network on PHP-Friends.
    Netcup and Hetzner are shared.

    PHP-Friends connects every physical server to 4x 10 GBit/s
    https://php-friends.de/vserver-ssd/hostsysteme
    And then gives you 1Gbps dedicated, its unheard of in this price range.

    You can have shared 10Gbps on PHP-Friends if you prefer that, they have couple of nodes configured that way. Schnupperspecial 2021 offer was 10Gbit shared for example, but this won't give you consistent speeds, just like Hetzner/netcup doesnt do it. Dedicated network will always be better if you want best reliability.

    In other words, Hetzner's cheapest vds has one 7763 cpu core, 4 gb ram, 100-200 mpbs 20 TB network, almost 100 mbps unmetered (32 tb). 7763 single thread passmark is 2600, one core 2 threads about 5000 passmark.

    In netcup you can use 120TB traffic and then it slowns down to 200Mbps.
    In PHP-Friends you can use 10TB traffic and then its 100Mbps (you can easily go above 33TB with paying zero extra).
    In Hetzner you use 20TB and then you pay 1euro/1TB so if you want to have 120TB like netcup then pay 100euro more.

    Hetzner's "almost 100 mbps unmetered" that you are most excited about is literally the worst of all.

    Both Netcup and php friends need to increase their network port to 10 gbps.

    Then they need to charge for bandwidth like Hetzner, make more limits or increase price.

  • letloverletlover Member

    @AXYZE said:

    @letlover said:
    Then each big server can host about 50-60 Hetnzer vds, 25-30 Netcup vds, 25-30 php friends vds.

    They can and they host more than that.

    If the big server is dual socket, then Hetzner's vds basic bandwidth is 10/100=100 mbps, Netcup 50 mbps, php friends 25 mbps.

    No. You have 1Gbps DEDICATED network on PHP-Friends.
    Netcup and Hetzner are shared.

    PHP-Friends connects every physical server to 4x 10 GBit/s
    https://php-friends.de/vserver-ssd/hostsysteme
    And then gives you 1Gbps dedicated, its unheard of in this price range.

    You can have shared 10Gbps on PHP-Friends if you prefer that, they have couple of nodes configured that way. Schnupperspecial 2021 offer was 10Gbit shared for example, but this won't give you consistent speeds, just like Hetzner/netcup doesnt do it. Dedicated network will always be better if you want best reliability.

    In other words, Hetzner's cheapest vds has one 7763 cpu core, 4 gb ram, 100-200 mpbs 20 TB network, almost 100 mbps unmetered (32 tb). 7763 single thread passmark is 2600, one core 2 threads about 5000 passmark.

    In netcup you can use 120TB traffic and then it slowns down to 200Mbps.
    In PHP-Friends you can use 10TB traffic and then its 100Mbps (you can easily go above 33TB with paying zero extra).
    In Hetzner you use 20TB and then you pay 1euro/1TB so if you want to have 120TB like netcup then pay 100euro more.

    Hetzner's "almost 100 mbps unmetered" that you are most excited about is literally the worst of all.

    Both Netcup and php friends need to increase their network port to 10 gbps.

    Then they need to charge for bandwidth like Hetzner, make more limits or increase price.

    Then php friends $8 vds is the best, beating the other two heavily.

  • AXYZEAXYZE Member
    edited June 16

    @letlover said:

    @AXYZE said:

    @letlover said:
    Then each big server can host about 50-60 Hetnzer vds, 25-30 Netcup vds, 25-30 php friends vds.

    They can and they host more than that.

    If the big server is dual socket, then Hetzner's vds basic bandwidth is 10/100=100 mbps, Netcup 50 mbps, php friends 25 mbps.

    No. You have 1Gbps DEDICATED network on PHP-Friends.
    Netcup and Hetzner are shared.

    PHP-Friends connects every physical server to 4x 10 GBit/s
    https://php-friends.de/vserver-ssd/hostsysteme
    And then gives you 1Gbps dedicated, its unheard of in this price range.

    You can have shared 10Gbps on PHP-Friends if you prefer that, they have couple of nodes configured that way. Schnupperspecial 2021 offer was 10Gbit shared for example, but this won't give you consistent speeds, just like Hetzner/netcup doesnt do it. Dedicated network will always be better if you want best reliability.

    In other words, Hetzner's cheapest vds has one 7763 cpu core, 4 gb ram, 100-200 mpbs 20 TB network, almost 100 mbps unmetered (32 tb). 7763 single thread passmark is 2600, one core 2 threads about 5000 passmark.

    In netcup you can use 120TB traffic and then it slowns down to 200Mbps.
    In PHP-Friends you can use 10TB traffic and then its 100Mbps (you can easily go above 33TB with paying zero extra).
    In Hetzner you use 20TB and then you pay 1euro/1TB so if you want to have 120TB like netcup then pay 100euro more.

    Hetzner's "almost 100 mbps unmetered" that you are most excited about is literally the worst of all.

    Both Netcup and php friends need to increase their network port to 10 gbps.

    Then they need to charge for bandwidth like Hetzner, make more limits or increase price.

    Then php friends $8 vds is the best, beating the other two heavily.

    Depends what you need.

    netcup has most traffic included + big drive + a lot of CPU power at low price.
    Hetzner has hourly billing & 10Gbit burst network.
    PHP-Friends has one of the best RAM/price ratio + best reliability because they dedicate most resources to your instance so it runs good 24/7.

    netcup would be great for machine learning/video encoding/video hosting/streaming videos, Hetzner would be great for cluster like kubernetes (easy & fast scaling, you dont need to pay for whole month/year because of hourly billing), PHP-Friends best for critical apps. I can also vouch for PHP-Friends support - because they are not that big they really care about you and will take their time to even finetune their DDoS protection. Hetzner won't care, they will say that you need to buy extra DDoS protection somewhere else.

    Thanked by 1letlover
  • letloverletlover Member

    I think php friends already breaks the bottleneck of lowend bmds. KS1 is only good for seed for its 2tb hdd. Otherwise, php friends vds is the way for low to mid traffic web hosting. Its passmark and ram are good enough for even JAVA based framework.

  • letloverletlover Member

    "For many years, we have been relying on Supermicro as a strong partner in the field of server hardware. For the current product generation (SSD G3), we purchase systems fully assembled by Supermicro. This not only gives us great flexibility, but also guarantees 100% compatibility of the components used. In this way, we ensure a long-term stable and high-performance operation. We only use our own hardware, which we take care of ourselves. A "resale" of purchased server resources does not take place on our part.

    Each host system has at least the following equipment:

    Processors: 2x Intel Xeon Gold (6140 / 6240) or 2x AMD EPYC 7452
    Memory: 1 TB DDR4 ECC RAM (Intel: 2666 MHz, AMD: 3200 MHz)
    Storage: 8 NVMe SSDs in RAID-10 (Samsung PM1725a / PM1725b or Micron 9300 MAX / PRO)
    Redundant uplinks: 4x 10 Gbps (connected to different switches via LACP)
    Redundant power supplies connected to two independent circuits
    Components such as fans, SSDs and power supplies are not only redundant, but can also be replaced in the unlikely event of a defect during operation. There is no impairment of customer systems."

    Supermicro mo, RAID 10 NVME, dedi 1 gbps port. Really good, comparable to lowend BMDS. If it has choice of HDD with large storage like 500 gb in its cheapest VDS, the application scenario will be much wider.

  • AXYZEAXYZE Member

    @letlover said:
    "For many years, we have been relying on Supermicro as a strong partner in the field of server hardware. For the current product generation (SSD G3), we purchase systems fully assembled by Supermicro. This not only gives us great flexibility, but also guarantees 100% compatibility of the components used. In this way, we ensure a long-term stable and high-performance operation. We only use our own hardware, which we take care of ourselves. A "resale" of purchased server resources does not take place on our part.

    Each host system has at least the following equipment:

    Processors: 2x Intel Xeon Gold (6140 / 6240) or 2x AMD EPYC 7452
    Memory: 1 TB DDR4 ECC RAM (Intel: 2666 MHz, AMD: 3200 MHz)
    Storage: 8 NVMe SSDs in RAID-10 (Samsung PM1725a / PM1725b or Micron 9300 MAX / PRO)
    Redundant uplinks: 4x 10 Gbps (connected to different switches via LACP)
    Redundant power supplies connected to two independent circuits
    Components such as fans, SSDs and power supplies are not only redundant, but can also be replaced in the unlikely event of a defect during operation. There is no impairment of customer systems."

    Supermicro mo, RAID 10 NVME, dedi 1 gbps port. Really good, comparable to lowend BMDS. If it has choice of HDD with large storage like 500 gb in its cheapest VDS, the application scenario will be much wider.

    They sell HDD as additional "backup space" which is located in nearby datacenter (couple kilometers away, connected by their own "dark fiber").
    Overall their offer is solid, they are geeks, linux admins for 10+ years, CEO himself responds to some support tickets if he has time... You really cant go wrong with them.

    Only con is their website design, but they just dont want the most flashy and they dont focus on marketing - they have their customers and they support them instead of trying to get more and more. Unusual tactic, but works great.

    This is #1 provider for me.

    Thanked by 2nick_ letlover
  • ArkasArkas Member, Moderator

    I feel that Hetzner is a little better than the others mentioned. Not by much, but it's that small difference that makes them appealing. Very flexible plans, fast support and network.

  • letloverletlover Member

    @AXYZE said:

    @letlover said:
    "For many years, we have been relying on Supermicro as a strong partner in the field of server hardware. For the current product generation (SSD G3), we purchase systems fully assembled by Supermicro. This not only gives us great flexibility, but also guarantees 100% compatibility of the components used. In this way, we ensure a long-term stable and high-performance operation. We only use our own hardware, which we take care of ourselves. A "resale" of purchased server resources does not take place on our part.

    Each host system has at least the following equipment:

    Processors: 2x Intel Xeon Gold (6140 / 6240) or 2x AMD EPYC 7452
    Memory: 1 TB DDR4 ECC RAM (Intel: 2666 MHz, AMD: 3200 MHz)
    Storage: 8 NVMe SSDs in RAID-10 (Samsung PM1725a / PM1725b or Micron 9300 MAX / PRO)
    Redundant uplinks: 4x 10 Gbps (connected to different switches via LACP)
    Redundant power supplies connected to two independent circuits
    Components such as fans, SSDs and power supplies are not only redundant, but can also be replaced in the unlikely event of a defect during operation. There is no impairment of customer systems."

    Supermicro mo, RAID 10 NVME, dedi 1 gbps port. Really good, comparable to lowend BMDS. If it has choice of HDD with large storage like 500 gb in its cheapest VDS, the application scenario will be much wider.

    They sell HDD as additional "backup space" which is located in nearby datacenter (couple kilometers away, connected by their own "dark fiber").
    Overall their offer is solid, they are geeks, linux admins for 10+ years, CEO himself responds to some support tickets if he has time... You really cant go wrong with them.

    Only con is their website design, but they just dont want the most flashy and they dont focus on marketing - they have their customers and they support them instead of trying to get more and more. Unusual tactic, but works great.

    This is #1 provider for me.

    Can their backup space hdd be used for website database for mounting?

  • AXYZEAXYZE Member

    @letlover said:

    @AXYZE said:

    @letlover said:
    "For many years, we have been relying on Supermicro as a strong partner in the field of server hardware. For the current product generation (SSD G3), we purchase systems fully assembled by Supermicro. This not only gives us great flexibility, but also guarantees 100% compatibility of the components used. In this way, we ensure a long-term stable and high-performance operation. We only use our own hardware, which we take care of ourselves. A "resale" of purchased server resources does not take place on our part.

    Each host system has at least the following equipment:

    Processors: 2x Intel Xeon Gold (6140 / 6240) or 2x AMD EPYC 7452
    Memory: 1 TB DDR4 ECC RAM (Intel: 2666 MHz, AMD: 3200 MHz)
    Storage: 8 NVMe SSDs in RAID-10 (Samsung PM1725a / PM1725b or Micron 9300 MAX / PRO)
    Redundant uplinks: 4x 10 Gbps (connected to different switches via LACP)
    Redundant power supplies connected to two independent circuits
    Components such as fans, SSDs and power supplies are not only redundant, but can also be replaced in the unlikely event of a defect during operation. There is no impairment of customer systems."

    Supermicro mo, RAID 10 NVME, dedi 1 gbps port. Really good, comparable to lowend BMDS. If it has choice of HDD with large storage like 500 gb in its cheapest VDS, the application scenario will be much wider.

    They sell HDD as additional "backup space" which is located in nearby datacenter (couple kilometers away, connected by their own "dark fiber").
    Overall their offer is solid, they are geeks, linux admins for 10+ years, CEO himself responds to some support tickets if he has time... You really cant go wrong with them.

    Only con is their website design, but they just dont want the most flashy and they dont focus on marketing - they have their customers and they support them instead of trying to get more and more. Unusual tactic, but works great.

    This is #1 provider for me.

    Can their backup space hdd be used for website database for mounting?

    As far as I know yes, but I didnt tried it. You can mount drive to VPS and host data there, but theres some latency so I would suggest to cache database in RAM - Query Cache in MySQL or add Redis/Memcached.

    Thanked by 1letlover
  • fanfan Member

    PHP-friends looks interesting, their last offer here was 500mbps dedicated, maybe someday they can post some 1gbps dedicated special here.

  • Netcup Root line is great, they are cheaper than the Hetzner standard line and they are stable. As for PHP Friends, I was a client of their promotional VPS on Xeon E5 and it's weak for me, although the price is very good, I got rid of it.

    but Netcup has been having problems with the route lately, Frankfurt-Nuremberg(DC) is going through Vienna

    ---------------------------------------------------------------------------
     OS           : Debian GNU/Linux 10 (64 Bit)
     Virt/Kernel  : KVM / 4.19.0-20-amd64
     CPU Model    : AMD EPYC 7702P 64-Core Processor
     CPU Cores    : 4 @ 1996.229 MHz x86_64 512 KB Cache
     CPU Flags    : AES-NI Enabled & VM-x/AMD-V Disabled
     Load Average : 0.12, 0.24, 0.17
     Total Space  : 315G (182G ~61% used)
     Total RAM    : 16042 MB (3104 MB + 12563 MB Buff in use)
     Total SWAP   : 4096 MB (494 MB in use)
     Uptime       : 5 days 15:24
    ---------------------------------------------------------------------------
     ASN & ISP    : AS197540, netcup GmbH
     Organization : netcup GmbH
     Location     : Karlsruhe, Germany / DE
     Region       : Baden-Württemberg
    ---------------------------------------------------------------------------
    
     ## Geekbench v4 CPU Benchmark:
    
      Single Core : 4358  (VERY GOOD)
       Multi Core : 13619
    
     ## IO Test
    
     CPU Speed:
        bzip2     : 117 MB/s
       sha256     : 208 MB/s
       md5sum     : 480 MB/s
    
     RAM Speed:
       Avg. write : 2560.0 MB/s
       Avg. read  : 6314.7 MB/s
    
     Disk Speed:
       1st run    : 952 MB/s
       2nd run    : 356 MB/s
       3rd run    : 618 MB/s
       -----------------------
       Average    : 642.0 MB/s
    
     ## Europe Speedtest.net
    
     Location                         Upload           Download         Ping   
    ---------------------------------------------------------------------------
     Nearby                           978.54 Mbit/s    1565.81 Mbit/s   20.086 ms
    ---------------------------------------------------------------------------
     UK, London (toob Ltd)            645.92 Mbit/s    1259.56 Mbit/s   15.761 ms
     Netherlands, Amsterdam (XS4ALL)  900.31 Mbit/s    1163.57 Mbit/s   11.344 ms
     Germany, Berlin (DNS:NET)        965.79 Mbit/s    1605.65 Mbit/s   14.239 ms
     Germany, Munich (InterNetX)      608.17 Mbit/s    1237.95 Mbit/s   10.214 ms
     Denmark, Copenhagen (Fiberby)    579.97 Mbit/s    600.15 Mbit/s    10.023 ms
     Sweden, Stockholm (SUNET)        371.93 Mbit/s    755.45 Mbit/s    28.273 ms
     Norway, Oslo (NextGenTel)        565.16 Mbit/s    1121.12 Mbit/s   22.271 ms
     France, Lyon (SFR)               191.04 Mbit/s    622.41 Mbit/s    22.504 ms
     Spain, Madrid (MasMovil)         420.12 Mbit/s    1149.66 Mbit/s   27.903 ms
     Italy, Rome (Unidata)            183.90 Mbit/s    589.76 Mbit/s    40.801 ms
     Czechia, Prague (Nordic Telecom) 900.96 Mbit/s    1776.17 Mbit/s   10.443 ms
     Austria, Vienna (A1)             711.56 Mbit/s    1405.53 Mbit/s   12.049 ms
     Poland, Warsaw (ISP Emitel)      542.54 Mbit/s    830.12 Mbit/s    20.886 ms
     Ukraine, Kyiv (KyivStar)         319.25 Mbit/s    1128.29 Mbit/s   40.034 ms
     Latvia, Riga (Bite)              289.63 Mbit/s    750.35 Mbit/s    35.120 ms
     Romania, Bucharest (DOTRO Tel)   128.55 Mbit/s    92.78 Mbit/s     26.288 ms
     Greece, Athens (GRNET)           406.90 Mbit/s    918.01 Mbit/s    41.699 ms
     Turkey, Urfa (Firatnet)          109.89 Mbit/s    281.85 Mbit/s    79.593 ms
    ---------------------------------------------------------------------------
    
     Finished in : 6 min 34 sec
     Timestamp   : 2022-06-16 18:06:21 GMT
     Saved in    : /root/speedtest.log
    
     Share results:
     - https://www.speedtest.net/result/13290691305.png
     - https://browser.geekbench.com/v4/cpu/16570699
     - https://clbin.com/lKeDg
    
  • ralfralf Member

    PHP friends sounds interesting. Do you happen to know if their "2 dedicated cores" means 2c/4t or 1c/2t? I'm guessing it's threads as they go up in 2s, but either way it still seems very good value compared to e.g. Hetzner CCX11.

    Ironically, I'd not even bothered looking at their offerings before because I hate PHP so much! :p

  • ArkasArkas Member, Moderator

    @ralf said: I hate PHP so much!

    What!!!!????
    I love PHP. It's the backbone of server side behavior (or behaviour depending where you are from) :blush:

  • letloverletlover Member

    I used to avoid anything php. But as Oracle charges Java SDK, I now avoid anything Java, although many of them are so good enterprise level mature web apps. Then Facebook improves php a lot. I guess with php 7 and php 8, the gap between php and python, ruby, perl, java, is not that significant. Now the true free open mature web frameworks are not many, php, python django, ror, per 6 maybe.

  • letloverletlover Member

    I feel Geekbench is more like American football. You get large difference in scores even the performance gap is very small between the two teams/cpus. Passmark is more reasonable, but not exciting, as you can see that not much has improved on cpu performance since 2011, except the energy efficiency of the cpu.

  • letloverletlover Member

    Another thing needs to be considered when comparing vds to bmds.
    Let's say, php friends L vds has 8 vcore, this is 4 cores 8 threads, basically a quadcore cpu. But this is withing a dual socket 32 core 64 threads hosting node server. Is this L vds totally equal to a single socket quad core cpu bmds?
    My gut feeling is no.
    There are two reasons.
    First, kvm, esix, or hypervisor creates a virtual layer, reduces the efficiency. In addition, these softwares may not totally optimized for the latest 64 core/32 core cpu dual socket servers, so parallel computing is less efficient than a single socket quadcore server.
    Second, all of the softwares on top of this virtual layer may not be optimized for parallel computing neither. This additionally causes further loss of efficiency.

    At the end of the day, though, 90% of the websites are low to mid low traffic, they can be hosted by php friends S vds without a problem, as long as php friends vds has some more hdd/ssd large volume solution, either attached volume in addition to the nvme system disk or direct 1 tb volume without nvme. I think php friends S vds with its two threads 5000 passmark 10 gb ram raid 10 nvme 1 gbps 10 tb bw is almost a high end bmds 10 years ago. Ten years ago dedicated 1 gbps port and 10 gb ram are very good features, and nvme is not populized yet (there are fusion disk on pcie 16x already, but very expensive).

  • @letlover said: Let's say, php friends L vds has 8 vcore, this is 4 cores 8 threads, basically a quadcore cpu. But this is withing a dual socket 32 core 64 threads hosting node server. Is this L vds totally equal to a single socket quad core cpu bmds?
    My gut feeling is no.

    Unless the single quad core cpu is supposed to be much faster, the 8 vcore can be faster if they are not pegged to 4 physical cores on a server that is not fully loaded.

    Honestly, I think you might be overthinking a lot of things in a bad way.

    Thanked by 1letlover
  • ralfralf Member

    @Arkas said:

    @ralf said: I hate PHP so much!

    What!!!!????
    I love PHP. It's the backbone of server side behavior (or behaviour depending where you are from) :blush:

    Yeah, it is. And I'm sure it's a lot nicer nowadays. But when I was using PHP4 about 20 years ago, it was already pretty popular, but encouraged terrible programming styles. I ended up maintaining support for a few legacy projects (so not really enough time to fix the issues, just the minimum to track down reported bugs and fix them) and I hated the spaghetti code I found everywhere.

    It's not really the language, you can get bad code everywhere, but after shifting to Java (back when it was still all about JavaBeans and JWS and Tomcat was just starting) the paradigm of separating things into model, view and controller was compelling, and instances of terrible code was much harder to find.

    I'm sure modern PHP encourages better programming styles, and of course it was always possible to do it nicely in PHP4, but the temptation was always to tightly couple business logic and presentation in the same file just to get something done quickly.

    So, it's not really rational, I just don't especially like PHP. That said, I used PHP for my personal homepage (last updated over 10 years ago) because it was the quickest way to solving the problem of having a consistent with just a single line include at the top of each page.

  • cadddrcadddr Member

    Does netcup enable nested virt now? Otherwise php-friends is a clear winner for me.

  • letloverletlover Member

    Another thing to consider is security.

    BMDS with one single service is the most secure. To cheap out, more services on one BMDS is normal, but this compromises the security. Still the end user can control which service is safe enough to coexist. VDS/VPS nodes have all kinds of customers with different levels of expertise and different interests. The only best protection is the virtual layer to separate each VDS/VPS. However, KVM or others like ESI Hypervisor are not immune to container hacker attack. So inherently VDS/VPS are more vulnerable than BMDS.

    At the price range of >$30, VDS cannot compete with Hetzner AX41 at all, except elastic computing, due to its no need for set up fee. At the price range of $10-
    $30, VDS has the latest hardware and software vs old hardware BMDS, which one to choose is up to the customer. Below $10, there is no real BMDS offer, VDS really shines here.

    PHP friends S VDS really nails it with its price tag and its computing power and THE dedicated 1 gbps. php friends may very well be the future Hetzner in the VDS business.

  • letloverletlover Member

    @AXYZE said:

    @letlover said:

    @AXYZE said:

    @letlover said:
    "For many years, we have been relying on Supermicro as a strong partner in the field of server hardware. For the current product generation (SSD G3), we purchase systems fully assembled by Supermicro. This not only gives us great flexibility, but also guarantees 100% compatibility of the components used. In this way, we ensure a long-term stable and high-performance operation. We only use our own hardware, which we take care of ourselves. A "resale" of purchased server resources does not take place on our part.

    Each host system has at least the following equipment:

    Processors: 2x Intel Xeon Gold (6140 / 6240) or 2x AMD EPYC 7452
    Memory: 1 TB DDR4 ECC RAM (Intel: 2666 MHz, AMD: 3200 MHz)
    Storage: 8 NVMe SSDs in RAID-10 (Samsung PM1725a / PM1725b or Micron 9300 MAX / PRO)
    Redundant uplinks: 4x 10 Gbps (connected to different switches via LACP)
    Redundant power supplies connected to two independent circuits
    Components such as fans, SSDs and power supplies are not only redundant, but can also be replaced in the unlikely event of a defect during operation. There is no impairment of customer systems."

    Supermicro mo, RAID 10 NVME, dedi 1 gbps port. Really good, comparable to lowend BMDS. If it has choice of HDD with large storage like 500 gb in its cheapest VDS, the application scenario will be much wider.

    They sell HDD as additional "backup space" which is located in nearby datacenter (couple kilometers away, connected by their own "dark fiber").
    Overall their offer is solid, they are geeks, linux admins for 10+ years, CEO himself responds to some support tickets if he has time... You really cant go wrong with them.

    Only con is their website design, but they just dont want the most flashy and they dont focus on marketing - they have their customers and they support them instead of trying to get more and more. Unusual tactic, but works great.

    This is #1 provider for me.

    Can their backup space hdd be used for website database for mounting?

    As far as I know yes, but I didnt tried it. You can mount drive to VPS and host data there, but theres some latency so I would suggest to cache database in RAM - Query Cache in MySQL or add Redis/Memcached.

    Does PHP Friends provide backup feature, like LightSail's snapshot function? This way if the website needs upgrade to a bigger vds instance, it is very easy to do with snapshot.

  • letloverletlover Member

    curl -s https://raw.githubusercontent.com/masonr/yet-another-bench-script/master/yabs.sh | bash

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Yet-Another-Bench-Script

    v2022-06-11

    https://github.com/masonr/yet-another-bench-script

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Sun Jun 19 09:01:54 CEST 2022

    Basic System Information:

    Uptime : 0 days, 2 hours, 14 minutes
    Processor : AMD EPYC 7452 32-Core Processor
    CPU cores : 2 @ 2349.998 MHz
    AES-NI : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM : 9.8 GiB
    Swap : 2.0 GiB
    Disk : 76.7 GiB
    Distro : Ubuntu 18.04.6 LTS
    Kernel : 4.15.0-187-generic

    fio Disk Speed Tests (Mixed R/W 50/50):

    Block Size 4k (IOPS) 64k (IOPS)
    Read 83.81 MB/s (20.9k) 1.20 GB/s (18.8k)
    Write 84.04 MB/s (21.0k) 1.21 GB/s (18.9k)
    Total 167.85 MB/s (41.9k) 2.41 GB/s (37.7k)
    Block Size 512k (IOPS) 1m (IOPS)
    ------ --- ---- ---- ----
    Read 2.69 GB/s (5.2k) 2.07 GB/s (2.0k)
    Write 2.83 GB/s (5.5k) 2.21 GB/s (2.1k)
    Total 5.53 GB/s (10.8k) 4.28 GB/s (4.1k)

    iperf3 Network Speed Tests (IPv4):

    Provider | Location (Link) | Send Speed | Recv Speed
    | | |
    Clouvider | London, UK (10G) | 6.86 Gbits/sec | 522 Mbits/sec
    Online.net | Paris, FR (10G) | 7.44 Gbits/sec | 496 Mbits/sec
    Hybula | The Netherlands (40G) | 7.79 Gbits/sec | 781 Mbits/sec
    Uztelecom | Tashkent, UZ (10G) | 2.02 Gbits/sec | 153 Mbits/sec
    Clouvider | NYC, NY, US (10G) | 1.80 Gbits/sec | 125 Mbits/sec
    Clouvider | Dallas, TX, US (10G) | 1.49 Gbits/sec | 96.6 Mbits/sec
    Clouvider | Los Angeles, CA, US (10G) | 1.22 Gbits/sec | 109 Mbits/sec

    iperf3 Network Speed Tests (IPv6):

    Provider | Location (Link) | Send Speed | Recv Speed
    | | |
    Clouvider | London, UK (10G) | 6.66 Gbits/sec | 8.84 Gbits/sec
    Online.net | Paris, FR (10G) | 7.70 Gbits/sec | 7.97 Gbits/sec
    Hybula | The Netherlands (40G) | 8.76 Gbits/sec | 7.38 Gbits/sec
    Uztelecom | Tashkent, UZ (10G) | 2.46 Gbits/sec | 2.35 Gbits/sec
    Clouvider | NYC, NY, US (10G) | 2.34 Gbits/sec | 2.37 Gbits/sec
    Clouvider | Dallas, TX, US (10G) | 1.50 Gbits/sec | 1.50 Gbits/sec
    Clouvider | Los Angeles, CA, US (10G) | 1.23 Gbits/sec | 1.30 Gbits/sec

    Geekbench 5 Benchmark Test:

    Test | Value
    |
    Single Core | 791
    Multi Core | 1537
    Full Test | https://browser.geekbench.com/v5/cpu/15546737

  • letloverletlover Member

    Just did some standard bench test on a S VDL of php friends.
    The OS is Ubuntu 18.04 LTS.
    I tried several times the bench test command. Only one test was complete and the results were shown above. All are good except the ipv4 receiving speed.

    I don't know why the bench test does not go through everytime. My one provider $7 and KS1 although do the bench test slowly, but very stably going through every step. Both are in Debian 10. Not sure if the OS causes the difference. Debian supposes to be very similar to Ubuntu.

  • letloverletlover Member

    curl -s https://raw.githubusercontent.com/masonr/yet-another-bench-script/master/yabs.sh | bash

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Yet-Another-Bench-Script

    v2022-06-11

    https://github.com/masonr/yet-another-bench-script

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Mon 20 Jun 2022 03:29:38 AM CEST

    Basic System Information:

    Uptime : 0 days, 0 hours, 7 minutes
    Processor : AMD EPYC 7452 32-Core Processor
    CPU cores : 2 @ 2349.998 MHz
    AES-NI : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM : 9.8 GiB
    Swap : 0.0 KiB
    Disk : 76.7 GiB
    Distro : Debian GNU/Linux 10 (buster)
    Kernel : 4.19.0-5-amd64

    fio Disk Speed Tests (Mixed R/W 50/50):

    Block Size 4k (IOPS) 64k (IOPS)
    Read 99.01 MB/s (24.7k) 1.21 GB/s (18.9k)
    Write 99.27 MB/s (24.8k) 1.22 GB/s (19.0k)
    Total 198.29 MB/s (49.5k) 2.43 GB/s (38.0k)
    Block Size 512k (IOPS) 1m (IOPS)
    ------ --- ---- ---- ----
    Read 2.32 GB/s (4.5k) 2.17 GB/s (2.1k)
    Write 2.45 GB/s (4.7k) 2.32 GB/s (2.2k)
    Total 4.77 GB/s (9.3k) 4.49 GB/s (4.3k)

    iperf3 Network Speed Tests (IPv4):

    Provider | Location (Link) | Send Speed | Recv Speed
    | | |
    Clouvider | London, UK (10G) | 8.41 Gbits/sec | 7.07 Gbits/sec
    Online.net | Paris, FR (10G) | busy | 7.09 Gbits/sec
    Hybula | The Netherlands (40G) | 8.28 Gbits/sec | 7.21 Gbits/sec
    Uztelecom | Tashkent, UZ (10G) | 2.23 Gbits/sec | 2.19 Gbits/sec
    Clouvider | NYC, NY, US (10G) | 2.22 Gbits/sec | 548 Mbits/sec
    Performing IPv4 iperf3 send test to Clouvider (Attempt #1 of 3)...client_loop: send disconnect: Broken pipe

    I reinstalled debian 10. The benchmark still is not successful. The error message is "Broken pipe"

  • letloverletlover Member

    I use root to install openvpn on debian 10. The vpn server does not work, due to something like dns issue. Probably something related to php friends' firewall setup. Or openvpn cannot use root to install? Anyways, I am going to set up it as a mailbox to check its ip rep. That is it.

  • letloverletlover Member

    I reinstalled ubuntu 18.04. Then installed openvpn. It works now. I guess the default dns server of the debian vpn is blocked by the hardware firewall somehow.
    Anyways, I used the vpn to play the online game. The ping is about 260-290, the variance is 20%, the same as KS1 and OP $7 bmds, much better than the Lightsail and Hetzner VPS with a variance of 100%. So php friends vds very possibly has a dedicated port, whether it is 1 gbps may need other kind of tests, but the stability is dedi port level. Cheers,

  • letloverletlover Member

    When using this vpn to access webhostingtalk, it is direct access, the same as my lightsail vpn. In comparison, my KS1 vpn has to go through the you are human captha test every time, and the OP atom vpn is blocked by wht. IP rep seems good.

  • letloverletlover Member

    Just don't know why the benchmark script does not go through reliably for network speed test. The IPV4 inbound speed seems very different for each benchmarking. Not sure why.

    Other than this issue, seems to me that this php friends s vds is a winner.

    I think that php friends may need to have the snapshot function and to add or change ip function for its vds, so that upgrading and flexibility and usability convenience will be much better.

  • @letlover said: The ping is about 260-290, the variance is 20%, the same as KS1 and OP $7 bmds, much better than the Lightsail and Hetzner VPS with a variance of 100%.

    Where are you from? What is your ping to the respective servers? There's a good chance the variance is your route to the servers and not their route to the game servers.

    @letlover said: I think that php friends may need to have the snapshot function and to add or change ip function for its vds, so that upgrading and flexibility and usability convenience will be much better.

    Probably not happening just because they use solusvm and that probably means they don't want to spend time developing anything. The other providers you mentioned like hetzner, netcup and lightsail made their own inhouse panels which is why they have snapshots.

  • AXYZEAXYZE Member
    edited June 20

    @letlover said:
    Just did some standard bench test on a S VDL of php friends.
    The OS is Ubuntu 18.04 LTS.
    I tried several times the bench test command. Only one test was complete and the results were shown above. All are good except the ipv4 receiving speed.
    I don't know why the bench test does not go through everytime.

    Its because DDoS mitigation kicked in. That's why you see reduced download speed (they inspect packets) and connetion drop. They have it configured to fit 99% of customers.

    You can message them to tweak protection if this affects your real life performance. :)

    PHP-Friends instances and networking are very hardened, they don't leave it "as-is". Like I said they have 10+ year experience in this stuff and tried to make most stable & reliable platform.

    Surprising that they are doing 10Gbps on their 1Gbps offer! It wasn't the case before.
    So right now its 1Gbps dedicated, 10Gbps burst, dedicated threads. For just 7.99euro/mo? Crazy deal if youre okay with 10/20/40TB traffic limit.

    You can see their traffic limit for every offer here https://php-friends.de/vserver-ssd/traffic

Sign In or Register to comment.