Howdy, Stranger!

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


Pulsed Media MiniServers // Transcoding Servers // Streaming Servers!! NVMe+1G Unmetered FROM 29.99€
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.

Pulsed Media MiniServers // Transcoding Servers // Streaming Servers!! NVMe+1G Unmetered FROM 29.99€

PulsedMediaPulsedMedia Member, Patron Provider
edited March 2023 in Offers

Pulsed Media MiniServers // Transcoding Servers // Streaming Servers!! NVMe+1G Unmetered FROM 29.99€

Pulsed Media MiniServers! i5-7500t / i7-7700t + 16/32G RAM + 250/2000GB NVMe + 1Gbps Unmetered!

These are the excellent new miniserver platform we have been working on! Pulling in all the experience from a decade of various small limited qty setups, this is it.
We have a clear path with these now to a truly massive scale.

What makes these miniservers truly spectacular is the built-in Intel Quick Sync for those of you who need hardware transcoding // streaming. On top of that, these have geekbench single core score above 900! But hey, i am certain someone here will post yabs results very shortly! ;)

We are still improving before putting these into massive scale, and once we get the new developments done we will upgrade the old nodes! :)

Dedicated Mini Servers

All units and more details at: https://pulsedmedia.com/minidedi-dedicated-servers-finland.php

Delivery within 2 business days.

Payment methods

  • Paypal
  • Crypto (BTC, LTC, XMR, ETH, DOGE and many others.)
«13456716

Comments

  • Really nice offer, wish you all best in sale.

  • Is that recurring deal?

    Thanked by 1greentea
  • PulsedMediaPulsedMedia Member, Patron Provider
    edited March 2023

    @Merlincool said:
    Is that recurring deal?

    That is standard recurring price yes.

    and these are super power efficient so there won't be any energy crisis risk with these neither.

    Thanked by 2greentea IHA_User
  • test ip or looking glass?

  • PulsedMediaPulsedMedia Member, Patron Provider

    @88xrzs said:
    test ip or looking glass?

    You can ping against a production server, say 185.148.1.70 (picked at random from that /24)

  • PulsedMediaPulsedMedia Member, Patron Provider

    YABS from one of the nodes:

    # ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
    #              Yet-Another-Bench-Script              #
    #                     v2023-02-27                    #
    # https://github.com/masonr/yet-another-bench-script #
    # ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
    
    Fri 03 Mar 2023 08:17:17 PM EET
    
    Basic System Information:
    ---------------------------------
    Uptime     : 9 days, 2 hours, 12 minutes
    Processor  : Intel(R) Core(TM) i5-7500T CPU @ 2.70GHz
    CPU cores  : 4 @ 2119.773 MHz
    AES-NI     : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM        : 15.5 GiB
    Swap       : 2.3 GiB
    Disk       : 1.8 TiB
    Distro     : Debian GNU/Linux 11 (bullseye)
    Kernel     : 5.10.0-18-amd64
    VM Type    : NONE
    
    Basic Network Information:
    ---------------------------------
    Protocol   : IPv4
    ISP        : Magna Capax Finland Oy
    ASN        : AS203003 Magna Capax Finland Oy
    Host       : Magnacapax
    Location   : Helsinki, Uusimaa (18)
    Country    : Finland
    
    fio Disk Speed Tests (Mixed R/W 50/50):
    ---------------------------------
    Block Size | 4k            (IOPS) | 64k           (IOPS)
      ------   | ---            ----  | ----           ---- 
    Read       | 475.67 MB/s (118.9k) | 633.06 MB/s   (9.8k)
    Write      | 476.92 MB/s (119.2k) | 636.39 MB/s   (9.9k)
    Total      | 952.60 MB/s (238.1k) | 1.26 GB/s    (19.8k)
               |                      |                     
    Block Size | 512k          (IOPS) | 1m            (IOPS)
      ------   | ---            ----  | ----           ---- 
    Read       | 805.23 MB/s   (1.5k) | 1.04 GB/s     (1.0k)
    Write      | 848.01 MB/s   (1.6k) | 1.11 GB/s     (1.0k)
    Total      | 1.65 GB/s     (3.2k) | 2.15 GB/s     (2.1k)
    
    iperf3 Network Speed Tests (IPv4):
    ---------------------------------
    Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping           
    -----           | -----                     | ----            | ----            | ----           
    Clouvider       | London, UK (10G)          | 823 Mbits/sec   | 927 Mbits/sec   | 36.0 ms        
    Scaleway        | Paris, FR (10G)           | 824 Mbits/sec   | 927 Mbits/sec   | 38.3 ms        
    NovoServe       | North Holland, NL (40G)   | 823 Mbits/sec   | 930 Mbits/sec   | 29.3 ms        
    Uztelecom       | Tashkent, UZ (10G)        | 725 Mbits/sec   | 877 Mbits/sec   | 109 ms         
    Clouvider       | NYC, NY, US (10G)         | 788 Mbits/sec   | 892 Mbits/sec   | 91.4 ms        
    Clouvider       | Dallas, TX, US (10G)      | busy            | busy            | 136 ms         
    Clouvider       | Los Angeles, CA, US (10G) | busy            | 191 Mbits/sec   | 166 ms         
    
    Geekbench 6 Benchmark Test:
    ---------------------------------
    Test            | Value                         
                    |                               
    Single Core     | 1245                          
    Multi Core      | 3737                          
    Full Test       | https://browser.geekbench.com/v6/cpu/394954
    
    YABS completed in 8 min 35 sec
    
    
  • yoursunnyyoursunny Member, IPv6 Advocate

    Do you have IPv6 and IPv9?

    Thanked by 1xms
  • PulsedMediaPulsedMedia Member, Patron Provider
    # ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
    #              Yet-Another-Bench-Script              #
    #                     v2023-02-27                    #
    # https://github.com/masonr/yet-another-bench-script #
    # ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
    
    Fri 03 Mar 2023 08:31:58 PM EET
    
    Basic System Information:
    ---------------------------------
    Uptime     : 10 days, 3 hours, 3 minutes
    Processor  : Intel(R) Core(TM) i7-7700T CPU @ 2.90GHz
    CPU cores  : 8 @ 3191.241 MHz
    AES-NI     : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM        : 31.2 GiB
    Swap       : 2.3 GiB
    Disk       : 1.8 TiB
    Distro     : Debian GNU/Linux 11 (bullseye)
    Kernel     : 5.10.0-18-amd64
    VM Type    : NONE
    
    Basic Network Information:
    ---------------------------------
    Protocol   : IPv4
    ISP        : Magna Capax Finland Oy
    ASN        : AS203003 Magna Capax Finland Oy
    Host       : Magnacapax
    Location   : Helsinki, Uusimaa (18)
    Country    : Finland
    
    fio Disk Speed Tests (Mixed R/W 50/50):
    ---------------------------------
    Block Size | 4k            (IOPS) | 64k           (IOPS)
      ------   | ---            ----  | ----           ---- 
    Read       | 462.13 MB/s (115.5k) | 642.39 MB/s  (10.0k)
    Write      | 463.35 MB/s (115.8k) | 645.77 MB/s  (10.0k)
    Total      | 925.48 MB/s (231.3k) | 1.28 GB/s    (20.1k)
               |                      |                     
    Block Size | 512k          (IOPS) | 1m            (IOPS)
      ------   | ---            ----  | ----           ---- 
    Read       | 838.62 MB/s   (1.6k) | 1.02 GB/s     (1.0k)
    Write      | 883.17 MB/s   (1.7k) | 1.09 GB/s     (1.0k)
    Total      | 1.72 GB/s     (3.3k) | 2.11 GB/s     (2.0k)
    
    iperf3 Network Speed Tests (IPv4):
    ---------------------------------
    Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping           
    -----           | -----                     | ----            | ----            | ----           
    Clouvider       | London, UK (10G)          | 819 Mbits/sec   | 927 Mbits/sec   | 36.3 ms        
    Scaleway        | Paris, FR (10G)           | 828 Mbits/sec   | busy            | 34.5 ms        
    NovoServe       | North Holland, NL (40G)   | 823 Mbits/sec   | 930 Mbits/sec   | 29.4 ms        
    Uztelecom       | Tashkent, UZ (10G)        | 782 Mbits/sec   | 881 Mbits/sec   | 110 ms         
    Clouvider       | NYC, NY, US (10G)         | 786 Mbits/sec   | 893 Mbits/sec   | 91.3 ms        
    Clouvider       | Dallas, TX, US (10G)      | 758 Mbits/sec   | 862 Mbits/sec   | 136 ms         
    Clouvider       | Los Angeles, CA, US (10G) | 142 Mbits/sec   | 493 Mbits/sec   | 166 ms         
    
    Geekbench 6 Benchmark Test:
    ---------------------------------
    Test            | Value                         
                    |                               
    Single Core     | 1448                          
    Multi Core      | 4697                          
    Full Test       | https://browser.geekbench.com/v6/cpu/395087
    
    YABS completed in 7 min 49 sec
    
    

    @yoursunny said: Do you have IPv6 and IPv9?

    IPv69!
    Sounds like fun!

    But we are boring and just provide you with IPv4

  • _cece_cece Member
    edited March 2023

    does a rescue system exist? And is the reinstall automatically done? Can the rdns be set (by the customer)?

  • PulsedMediaPulsedMedia Member, Patron Provider

    @_cece said:
    does a rescue system exists? And is the reinstall automatically done?

    Yes and no. Remote reboots are still unconfigured (temp hw for that installed tho), but otherwise both of those are automated.

    Not sure we want to spend too much time on this temp remote reboot hardware since new one for next iteration of development has already been designed, and i think most of the parts too.

  • yoursunnyyoursunny Member, IPv6 Advocate

    No IPv6 hall of shame

    • GitHub, Globalping
    • Outlook
    • SparkPost
    • ColonCrossing - VirMach, RackNerd, CloudServer, HostNamaste (Dallas), cheapwindowsvps
    • XetHost
    • SoftShellWeb (San Jose only)
    • KhanWebHost (Turkey etc)
    • Wishosting
    • Reprise Hosting
    • HostWebis
    • Oplink
    • Leapswitch Networks (except Mumbai)
    • DedicatServer gardens
    • HostTiger HostTigger
    • ForwardWeb
    • SKB Enterprise
    • HostHatch (Sydney etc)
    • Watta Server, ONeil Online
    • HostMayo
    • Private-Hosting di Cipriano Oscar
    • GardenCloud gardens
    • CoreHosting
    • Hye Cloud
    • iHostART (OpenVZ)
    • VirmAche, Gullo New York
    • HostVDS
    • FireVPS
    • Rabisu
    • MyHBD
    • RackEdge
    • Hostaris
    • PulsedMedia

    Include IPv6 for no extra cost on every plan in every location to get delisted.

    Thanked by 4lll xms pangkus Alyx
  • PulsedMediaPulsedMedia Member, Patron Provider
    # ping tur-fi.node.route48.org -4
    PING  (185.218.193.178) 56(84) bytes of data.
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=1 ttl=56 time=5.17 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=2 ttl=56 time=5.27 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=3 ttl=56 time=5.25 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=4 ttl=56 time=5.03 ms
    ^C
    ---  ping statistics ---
    4 packets transmitted, 4 received, 0% packet loss, time 3004ms
    rtt min/avg/max/mdev = 5.026/5.178/5.266/0.094 ms
    
  • yoursunnyyoursunny Member, IPv6 Advocate

    @PulsedMedia said:
    ping tur-fi.node.route48.org -4

    If the tunnel is configured on the hypervisor / router, it counts as IPv6 being provided.
    If the tunnel is configured inside the VM / dedicated server, it does not count.
    Moreover, Route48 isn't DMCA ignored and doesn't allow high bandwidth usage.

  • crunchbitscrunchbits Member, Patron Provider, Top Host

    @PulsedMedia said:

    # ping tur-fi.node.route48.org -4
    PING  (185.218.193.178) 56(84) bytes of data.
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=1 ttl=56 time=5.17 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=2 ttl=56 time=5.27 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=3 ttl=56 time=5.25 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=4 ttl=56 time=5.03 ms
    ^C
    ---  ping statistics ---
    4 packets transmitted, 4 received, 0% packet loss, time 3004ms
    rtt min/avg/max/mdev = 5.026/5.178/5.266/0.094 ms
    

    Don’t give into terrorist demands. IPv4 for life.

    Nice offers, pretty neat approach especially with the energy issues. I like the hardware-as-a-solution first method.

  • yoursunnyyoursunny Member, IPv6 Advocate

    @crunchbits said:
    Don’t give into terrorist demands.

    https://lowendtalk.com/discussion/comment/3229065/#Comment_3229065

    @seriesn said:

    @jsg said:

    @yoursunny said:
    I won't name the provider, not even via PM, because I don't want them to break my legs.

    Break your legs?! Are we talking about a deranged provider or about a crime gang?

    Basically what @yoursunny does, if a provider doesn't offer IPv6.

    Thanked by 1crunchbits
  • crunchbitscrunchbits Member, Patron Provider, Top Host

    @yoursunny said:

    @crunchbits said:
    Don’t give into terrorist demands.

    https://lowendtalk.com/discussion/comment/3229065/#Comment_3229065

    @seriesn said:

    @jsg said:

    @yoursunny said:
    I won't name the provider, not even via PM, because I don't want them to break my legs.

    Break your legs?! Are we talking about a deranged provider or about a crime gang?

    Basically what @yoursunny does, if a provider doesn't offer IPv6.

    Are you threatening to involucrate me starting with my legs!?!

  • yoursunnyyoursunny Member, IPv6 Advocate

    @crunchbits said:
    Are you threatening to involucrate me starting with my legs!?!

    You did it to me two months ago.

    Thanked by 1crunchbits
  • crunchbitscrunchbits Member, Patron Provider, Top Host
    edited March 2023

    @yoursunny said:

    @crunchbits said:
    Are you threatening to involucrate me starting with my legs!?!

    You did it to me two months ago.

    Good, skipping leg day is like not offering IPv6.

    Thanked by 1yoursunny
  • PulsedMediaPulsedMedia Member, Patron Provider

    @crunchbits said:

    @PulsedMedia said:

    # ping tur-fi.node.route48.org -4
    PING  (185.218.193.178) 56(84) bytes of data.
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=1 ttl=56 time=5.17 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=2 ttl=56 time=5.27 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=3 ttl=56 time=5.25 ms
    64 bytes from 185.218.193.178 (185.218.193.178): icmp_seq=4 ttl=56 time=5.03 ms
    ^C
    ---  ping statistics ---
    4 packets transmitted, 4 received, 0% packet loss, time 3004ms
    rtt min/avg/max/mdev = 5.026/5.178/5.266/0.094 ms
    

    Don’t give into terrorist demands. IPv4 for life.

    Nice offers, pretty neat approach especially with the energy issues. I like the hardware-as-a-solution first method.

    :D :D Never ;)

    Regardless, if @yoursunny is happy to donate to us an 150k € router every ~5 years or so we can add full IPv6 routing table (~same size as IPv4 i think already ... going to be ^4 as large eventually), i guess we can then put the manpower to add IPv6 as an option for his dedi with us.... oh wait does @yoursunny have any service with us?

  • Such a good forum topic!

    Dear @PulsedMedia these are really great offers, don't have a need right now, but good to see you're doing your best to provide some really good and affordable stuff. Still, @_cece has some good points, give us customers way to reinstall, reboot, fix, repartition etc... This is important part of the service, IMHO.

    Dear @yoursunny I really, REALLY admire your efforts to boost IPv6 popularity, keep blacklists of providers stuck in the past, etc... Somebody had to do it, I'm glad we have you on the task! Having said that, while I consider your list, I must admit that I'm an oldskul guy, which even if provided IPv6 range, first thing I do is strip it, yeah, IPv4 only works for me the best (no need for ip6tables, ip6 nothing). :) But, still, thank you for your effort, I do consult your list to understand which providers are lazy or incompetent, actually. So your work is appreciated.

    Finally, @crunchbits thank you for really great service you provide to me, I like it a lot. But, having an open ticket with you for days, with no estimate when it could be done, and seeing you so active on the forum, I'm starting to think I won't be serviced properly (or soon) only because you waste so much time on the forum instead of tending to tickets. :smile: Please, make me wrong. :smile:

    There, I said it all, this beer is good and strong. :wink:

    Good luck with sales pulsed!

    Thanked by 1PulsedMedia
  • PulsedMediaPulsedMedia Member, Patron Provider

    @maverick said: Still, @_cece has some good points, give us customers way to reinstall, reboot, fix, repartition etc... This is important part of the service, IMHO.

    These are on the works, but we start with fundamentals first, on a as-needed basis. No fundamentals == Nothing to use all that automation for.

    This is kinda special setup, we have shared some info @ discord, i guess a blog post is in order after next iteration! :)

    @maverick said: which providers are lazy or incompetent

    Interestingly, it often is neither. I talk with many network operators, and a IP transit provider has been lately instructing some of their customers to remove IPv6 support when they had networking issues.

    One of them was because IPv6 routing table has already become so large that their router was dropping routes but no error was generated. IPv6 routes take 4x the memory in FIB. It is literally a money question for them, spend 100k €+ for a new router or just drop something 0% requires, and adds more functionality to only 1% of traffic?

    @maverick said: There, I said it all, this beer is good and strong. :wink:

    thumbs up

    @maverick said: Good luck with sales pulsed!

    Thanks

    Thanked by 2maverick lll
  • crunchbitscrunchbits Member, Patron Provider, Top Host

    @maverick said:
    Such a good forum topic!

    Dear @PulsedMedia these are really great offers, don't have a need right now, but good to see you're doing your best to provide some really good and affordable stuff. Still, @_cece has some good points, give us customers way to reinstall, reboot, fix, repartition etc... This is important part of the service, IMHO.

    Dear @yoursunny I really, REALLY admire your efforts to boost IPv6 popularity, keep blacklists of providers stuck in the past, etc... Somebody had to do it, I'm glad we have you on the task! Having said that, while I consider your list, I must admit that I'm an oldskul guy, which even if provided IPv6 range, first thing I do is strip it, yeah, IPv4 only works for me the best (no need for ip6tables, ip6 nothing). :) But, still, thank you for your effort, I do consult your list to understand which providers are lazy or incompetent, actually. So your work is appreciated.

    Finally, @crunchbits thank you for really great service you provide to me, I like it a lot. But, having an open ticket with you for days, with no estimate when it could be done, and seeing you so active on the forum, I'm starting to think I won't be serviced properly (or soon) only because you waste so much time on the forum instead of tending to tickets. :smile: Please, make me wrong. :smile:

    There, I said it all, this beer is good and strong. :wink:

    Good luck with sales pulsed!

    Depends what your ticket is, you can DM me instead of this thread. It’s possible it was mismarked, or possible there isn’t an answer ready for you—that won’t preclude me (or any support staff) from doing something else besides staring at your ticket.

    Also if you’re the one asking us for a /24 I already said no.

  • yoursunnyyoursunny Member, IPv6 Advocate
    edited March 2023

    @PulsedMedia said:
    One of them was because IPv6 routing table has already become so large that their router was dropping routes but no error was generated. IPv6 routes take 4x the memory in FIB. It is literally a money question for them, spend 100k €+ for a new router or just drop something 0% requires, and adds more functionality to only 1% of traffic?

    Full table fits in less than 512 MiB RAM:

    sunny@ixp2:~$ sudo birdc
    BIRD 2.0.7 ready.
    bird> show route count
    6 of 6 routes for 6 networks in table master4
    390402 of 390402 routes for 170603 networks in table master6
    70007 of 70007 routes for 70007 networks in table rpki6
    Total: 460415 of 460415 routes for 240616 networks in 3 tables
    bird> show memory
    BIRD memory usage
    Routing tables:     50 MB
    Route attributes:   62 MB
    Protocols:          97 kB
    Total:             113 MB
    
    Thanked by 1xms
  • PulsedMediaPulsedMedia Member, Patron Provider

    @yoursunny said:

    @PulsedMedia said:
    One of them was because IPv6 routing table has already become so large that their router was dropping routes but no error was generated. IPv6 routes take 4x the memory in FIB. It is literally a money question for them, spend 100k €+ for a new router or just drop something 0% requires, and adds more functionality to only 1% of traffic?

    Full table fits in less than 512 MiB RAM:

    sunny@ixp2:~$ sudo birdc
    BIRD 2.0.7 ready.
    bird> show route count
    6 of 6 routes for 6 networks in table master4
    390402 of 390402 routes for 170603 networks in table master6
    70007 of 70007 routes for 70007 networks in table rpki6
    Total: 460415 of 460415 routes for 240616 networks in 3 tables
    bird> show memory
    BIRD memory usage
    Routing tables:     50 MB
    Route attributes:   62 MB
    Protocols:          97 kB
    Total:             113 MB
    

    We use hardware routers only, RAM consumption is irrelevant. Any router can ingress magical numbers of routes, but what matters is what's built into the ASIC, Juniper calls this FIB.

    So how many CPUs you know with 226MiB+ of L1 cache?
    Then account for CPUs are mass market products, routers are not. Then you realize why a 4x100G ports can cost 35 000+€ ...

    Thanked by 1maverick
  • PulsedMediaPulsedMedia Member, Patron Provider
    edited March 2023

    further routers need "L3 cache" for packet buffers etc. it all becomes super expensive super fast.

    I think our current one has 384MiB or 512MiB PER PORT in the damn asic lol

  • @PulsedMedia said:

    These are on the works, but we start with fundamentals first, on a as-needed basis. No fundamentals == Nothing to use all that automation for.

    OK, I understand. Hit the market first, deal with the details later. I'll be patient, but look forward to get something from you later on. Would be my first gig in Finland, and I don't know anybody better than you. B)

    Interestingly, it often is neither. I talk with many network operators, and a IP transit provider has been lately instructing some of their customers to remove IPv6 support when they had networking issues.

    One of them was because IPv6 routing table has already become so large that their router was dropping routes but no error was generated. IPv6 routes take 4x the memory in FIB. It is literally a money question for them, spend 100k €+ for a new router or just drop something 0% requires, and adds more functionality to only 1% of traffic?

    One thing I like about you is that you're willing to honestly share some details, I remember that from other topics. Just another reason to consider having service with you. Thank you for these info! And my deepest apologies to providers that I wrongly called lazy, just because they couldn't spend another 100k € to keep those pesky IPv6 routing tables in RAM. :smile:

    Depends what your ticket is, you can DM me instead of this thread. It’s possible it was mismarked, or possible there isn’t an answer ready for you—that won’t preclude me (or any support staff) from doing something else besides staring at your ticket.

    Nah, your service so good I lost all links to the control panel. :wink: I'll keep waiting for an email from your support to recover links. :wink: Still, let's leave this topic to Pulsed, wanna see those all sold out first, then we'll get back to all famous US west coast storage (or in another topic). 200 ms from here, working great, a miracle. :smile:

    Thanked by 1crunchbits
  • PulsedMediaPulsedMedia Member, Patron Provider

    One more nitpick info, our current edge router has FOUR 2.2kw power supplies ... or was it 6, i forgot. Regardless, when we first booted it up we already measured kilowatts of power being used, and that was with limited linecards, zero usage.
    Supports 16 linecards.

    Each linecard has custom asic, so does the management cards. I think the empty chassis brand new (empty chassis == fans + power supplies + maybe management cardS) was about 50k €, and still one of the smaller ones.

    These routers go upto 1 full rack in size. They are delivered as a rack even.

    These are insane, the fans on these are by far the loudest in our DC. Everytime they change pitch you get a little bit scared of what happened, did something break? Nawh, ambient just changed by 1C ... :D

    Thanked by 1maverick
  • PulsedMediaPulsedMedia Member, Patron Provider

    @maverick said: OK, I understand. Hit the market first, deal with the details later. I'll be patient, but look forward to get something from you later on. Would be my first gig in Finland, and I don't know anybody better than you. B)

    You'd be surprised the things we are considering! ;) Let's hope many of them comes to fruition.

    Ultimately for dedis, those are limited usefulness because these are supposed to run without an hitch for years at a time.

    @maverick said: One thing I like about you is that you're willing to honestly share some details, I remember that from other topics. Just another reason to consider having service with you. Thank you for these info! And my deepest apologies to providers that I wrongly called lazy, just because they couldn't spend another 100k € to keep those pesky IPv6 routing tables in RAM. :smile:

    check what i just posted as well.

    If you buy brand new, routers are insanely expensive, and the funny thing?, that's the cheap part oO; It's the human labor.

    Example 1) Add IPv6, now you need to systematically map them to IPv4. IPv4 is easy to type and remember, IPv6 you most of the time need to copy paste just too long and too many variables many times.

    Example 2) Server with just IPv6 is useless. Server with IPv4 can do everything server with IPv6 can. BUT IPv6 forces you to use notepads, copy pasting, and extra documentation. IPv4 you just type and remember outright. I remember all our range first 3 segments outright, not a problem.

    RAM, I WISH it was RAM where all that info was stored, but FIB is comparable to L1/L2 cache in a CPU. It's SRAM built into the chip, very very close to the switching cores. It's not even just SRAM i believe, but a few transistors more to make the routing faster, latency is very key in these systems (hence why i advocate for going for larger packets!). I'm not very knowledgeable on this subject tho, on the fine points of the router ASIC microarchitecture, so take this with grain of salt.

    @maverick said: Still, let's leave this topic to Pulsed, wanna see those all sold out first,

    Btw, first place notified that these came into regular production schedule is LET :)

    Thanked by 1maverick
  • coventcovent Member
    edited March 2023

    @PulsedMedia nice to see reasonably priced dedi offerings from around here (other than classics Hetzner, Creanova...) :)

    Also good job on finding a sweet spot for those 2TB NvMEs, truly not many providers/plans around in this price range (with fully dedicated resources), I'd guess you'll sell a lot of these once the word gets out in wider scale

    1) do these prices already include VAT?
    2) can you give any hint about "new developments" you mentioned, you mean these given specs will get some boost still?

  • yoursunnyyoursunny Member, IPv6 Advocate

    @PulsedMedia said:
    We use hardware routers only, RAM consumption is irrelevant. Any router can ingress magical numbers of routes, but what matters is what's built into the ASIC, Juniper calls this FIB.

    So how many CPUs you know with 226MiB+ of L1 cache?
    Then account for CPUs are mass market products, routers are not. Then you realize why a 4x100G ports can cost 35 000+€ ...

    It's unnecessary to store whole FIB in SRAM.
    Instead, put the FIB in cheap DRAM and use SRAM as cache.

    Garegin Grigoryan and Yaoqing Liu. 2018. PFCA: a programmable FIB caching architecture. In Proceedings of the 2018 Symposium on Architectures for Networking and Communications Systems (ANCS '18). Association for Computing Machinery, New York, NY, USA, 97–103. https://doi.org/10.1145/3230718.3230721

Sign In or Register to comment.