Results 1 to 5 of 5
  1. #1

    PeeringBench - First ever network peering quality test

    Hi guys,

    I posted this on another forum, but wanted to share my website here as well in case someone missed it

    ------------------

    Long time lurker here and user of MANY low end boxes - usually dedicated servers in my case. Was never a fan of VPS...

    Anyways, for a long time I had a suspicion that the things we know about the "network quality" of certain providers shouldn't be taken at face value. You know, sometimes providers that cost more than the others, don't always offer better peering - especially over longer routes.

    So, since I had 4 of the big ones - Online, SyS (OVH), Hetzner and LeaseWeb at the same time, I decided to run a heck of a lot of different peering, latency and bandwidth benchmarks in order to more offer a better view of which providers is better at specific areas.

    My website is PeeringBench.com (it's a bit long but I believe if you're shopping for a new box, it's definitely worth a read).

    Would love to hear if my research helps any of you, and feel free to offer any suggestions!

    If everything goes well, I would love to test more boxes, like from RedStation, FDC, Server.Lu, etc., etc.

    ----------------

    Thanks!

  2. #2
    Join Date
    Nov 2005
    Posts
    3,944
    Not to rain on your parade but I wouldn't say speedtest.net is the most reliable method of testing the speed of gigabit servers. It can give you some insight, but isn't useful for true comparisons.

    1. Many big hosts have their own speed test servers. So you may be testing LAN rather than internet speeds.
    2. Many speed test servers are gigabit connections themselves, so if you have multiple people downloading at the same time, it isn't going to be a reliable result.


  3. #3
    Join Date
    Oct 2009
    Location
    Canada
    Posts
    485
    Completely agree with devonblzx. Furthermore, if you want to test "global" connectivity, you should be testing each server for bandwidth and latency performance from multiple points around the globe. I don't see how a single speedtest result allows you to conclude that the servers you tested have good global connectivity.

    For example: http://www.peeringbench.com/speedtests/SyS1.png. This test is with an OVH server, to an OVH speedtest.net server. Don't see how that measures anything but performance internal to OVH's network.
    █ Pentester & IT Security Consultant

  4. #4
    Quote Originally Posted by devonblzx View Post
    Not to rain on your parade but I wouldn't say speedtest.net is the most reliable method of testing the speed of gigabit servers. It can give you some insight, but isn't useful for true comparisons.

    1. Many big hosts have their own speed test servers. So you may be testing LAN rather than internet speeds.
    2. Many speed test servers are gigabit connections themselves, so if you have multiple people downloading at the same time, it isn't going to be a reliable result.
    I do agree with your opinion, although my website was never meant to be fully "scientific" research editorial - I just merely wanted some more concrete evidence and a better comparison between the "Big 4", at least better in quality than what most people usually read on forums.

    Do you know how many times people talk with a 100% belief that X network is better than Y or Z, without any kind of evidence? At least with these benchmarks, the opinions stated could at least be partially backed or disproved. I don't have a budget for doing more than this, and it's a completely free website.

    The torrent results for example use hundreds of different ASN's, so I do believe that it tests the providers network connectivity on a global level. Plus, all testings were done at around the same time, so the global web's congestion would still be the same for all providers.

    Though I still appreciate the opinion

    Quote Originally Posted by media-hosts_com View Post
    Completely agree with devonblzx. Furthermore, if you want to test "global" connectivity, you should be testing each server for bandwidth and latency performance from multiple points around the globe. I don't see how a single speedtest result allows you to conclude that the servers you tested have good global connectivity.

    For example: http://www.peeringbench.com/speedtests/SyS1.png. This test is with an OVH server, to an OVH speedtest.net server. Don't see how that measures anything but performance internal to OVH's network.
    It can still pinpoint to congestion inside a providers datacenter. Hetzner always delivered close to 1Gb internal connection, while OVH managed to get to 200-300Mb, which isn't positive for the nodes outside their datacenter, if they can't even achieve local gigabit speeds.

    Btw, if you scroll down the website you'll see many different points of testing, so the servers were indeed tested from dozens of points around the globe.

    Thanks!
    PeeringBench.com - First ever dedicated server peering and network quality benchmark!

  5. #5
    Join Date
    Nov 2005
    Posts
    3,944
    Quote Originally Posted by PeeringBench View Post
    It can still pinpoint to congestion inside a providers datacenter. Hetzner always delivered close to 1Gb internal connection, while OVH managed to get to 200-300Mb, which isn't positive for the nodes outside their datacenter, if they can't even achieve local gigabit speeds.
    Can it though? We're back to my original point. What if Hetzner had a 10gigabit connection on their speed test server and OVH had a gigabit connection? Three people testing at the same time on OVH would limit you to ~300mbps, while three people testing at the same time on Hetzner should still allow you to max out your port.

    There are a lot of variables and if you're going to run a site like this, if you're going to do it, make sure you do your research on this type of stuff and just don't rely on random speed test servers. Also, you need to have something in place to avoid congestion on the speed server itself and not the local server. What if 6 people were running the same test on Vodaphone while you were testing soyoustart and only 1 person was running the test when you were testing Hetzner. Did you run these tests at the same exact time? Did you run these tests at multiple times throughout the day and pick the best one, etc, etc, etc.
    Last edited by devonblzx; 07-27-2015 at 01:44 PM.


Similar Threads

  1. First ever host provider...?
    By dav in forum Web Hosting
    Replies: 1
    Last Post: 03-24-2004, 07:25 PM
  2. Do you remember your first ever webpage?
    By saj in forum Web Hosting Lounge
    Replies: 43
    Last Post: 10-27-2003, 12:16 AM
  3. First ever "Welcome Back!" domain sale
    By vipe in forum Other Offers & Requests
    Replies: 1
    Last Post: 02-16-2003, 01:34 PM
  4. Rack Shack network speed, quality of service and DNS
    By kickster in forum Dedicated Server
    Replies: 3
    Last Post: 10-22-2001, 04:37 AM
  5. first ever telnet session ends in disaster
    By heymatty in forum Dedicated Server
    Replies: 7
    Last Post: 08-07-2001, 02:03 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •