Howdy, Stranger!

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

Advertise on LowEndTalk.com
A Different way of benchmarking some popular hosts...
New on LowEndTalk? Please read our 'Community Rules' by clicking on it in the right menu!

A Different way of benchmarking some popular hosts...

NomadNomad Member
edited January 2015 in General

OK, So I'm trying to write a bash script to install Nginx with OpenResty, Nginx modules FFMpeg, ImageMagick, Php 5.5, MariaDB and Nginx tweaks & configurations... Et cetera...



The script installs most of the stuff from the source, except DotDeb Php and MariaDB.

I included a timer function to calculate how much time is spent (minus the time you spend on setting a mysql password page, I always forget to check that part so I is not calculated).

I ran this installation on some different providers I could lay my hand on.

The results are as below.

1- Vultr.com AMS 768 Mb RAM
          0 hours 27 minutes and 11 seconds

2- Runabove - Steadfast Res. L 4 Gb Ram 8 Cores
          0 hours 32 minutes and 52 seconds

3- DO AMS2 512MB Ram
          0 hours 40 minutes and 17 seconds

4- Linode London 512 Mb Ram
          0 hours 41 minutes and 17 seconds

5- Mirantis OpenStacks 4GB Ram 4 Cores
          0 hours 46 minutes and 0 seconds

6- DO NYC 512 Mb Ram

          0 hours 48 minutes and 14 seconds

7-UltraVPS.eu 1GB Ram

          0 hours 58 minutes and 41 seconds

8- Kimsufi KS-1 (N2800 one)

          2 hours 22 minutes and 7 seconds

All tests are done with a Debian 7 system.

The criteria for DO, Vultr and Linode was the cheapest package available.

As for RunAbove and Mirantis I chose the best I could get.

Kimsufi? Well... What was I expecting?

I do have an Online.net SCG-2 but I couldn't add it to tests since it's in use. If someone can and share the results I'ld be appreciated.




Strangely, every time I ran, Vultr finished way faster than all the rest.

OK, in my so-called-benchmark mostly the CPU power is important.

But that also means something...

Anyway, I just wanted to share.

If anyone wants to contribute a different result, it's appreciated. [Script]Note: Only for Debian

^_^

...
...

Comments

  • Vultr CPU performance is quite impressive.

  • I tested the script but did not see any results at the end of installation. Screen terminated immediately after DB & root password.

  • NomadNomad Member
    edited January 2015

    @century1stop What's the host? Debian 7? Can you check if you have a Time.log file in your home directory?
    Maybe I shouldn't use screen and I should let the text flow on the screen. Debugging is not easy this way. :(

    @kcaj Yeah, I wasn't expecting that much....

    ...
    ...

  • Tried on own server, will check time.log

  • century1stop, there was a locale problem with your server setup.
    Locale was not getting set, even manually. Strange.

    Thus MariaDB was giving an error, causing the script to terminate.
    I'm not familiar enough with OpenVZ systems but..
    The script terminated after 29 minutes 33 seconds. What was left is installing MariaDB, Php 5.5, Nginx configs and some little tweaks. It wouldn't take more than 1-2 minutes. So all in all.. Seems good.

    But seriously, I think I must test it with other OpenVZ and other systems as much as possible.

    ...
    ...

  • FlorisFloris Member
    edited January 2015

    Testing a Runabove Power 8 2XL instance with 48 GB of Ram and 176 IBM vCPU's and a Cloud Sandbox L with 4GB ram and 1 vCPU right now.

    Never mind, the 2XL instance failed running it (Ubuntu 14.04)

    I only work for me, and myself.

  • Yeah, Ubuntu will fail. I didn't set this up for Ubuntu. The dependencies are missing.

    Use Debian 7.
    ^^

    ...
    ...

  • Try runabove sandbox? the $2.50 2GB KVM

  • I just tried Steadfast, If you want, be my guest. (:

    ...
    ...

  • FlorisFloris Member
    edited January 2015

    @Nomad said:
    Yeah, Ubuntu will fail. I didn't set this up for Ubuntu. The dependencies are missing.

    Use Debian 7.
    ^^

    Could you adapt the script to support Fedora 17 or Ubuntu 14.04? I'd love to give that beast a spin. They claim it to be "up too 100 times faster"

    I only work for me, and myself.

  • @JoeMerit said:
    Try runabove sandbox? the $2.50 2GB KVM

    Already running the 4GB version.

    I only work for me, and myself.

  • I can, to Ubuntu at least. But not today- tonight. Tomorrow I'll give it a go. ;)

    ...
    ...

  • @Floris said:
    Could you adapt the script to support Fedora 17 or Ubuntu 14.04? I'd love to give that beast a spin. They claim it to be "up too 100 times faster"

    Which beast is "up to 100 times faster" ?

    ...
    ...

  • FlorisFloris Member
    edited January 2015

    @Nomad said:

    The RunAbove Power8 series. They claim their CPU cores are up too 100 times faster than a classic x86 setup

    I only work for me, and myself.

  • @Floris said:
    The RunAbove Power8 series. They claim their CPU cores are up too 100 times faster than a classic x86 setup

    I found their power8 to be slower than a 1-core vultr vps, here's a serverbear bench that I did: http://serverbear.com/benchmark/2014/12/29/YKJ5inaqZa1AWgiA

    tsdns.io - free, redundant, DDoS-protected TSDNS

  • I think power8 is something designed for multi threaded applications, not everything will be faster than x86.

  • Is Linode offering 512MB instance?

  • @Maniac Upps, copy/paste error. 1024 it is.

    ...
    ...

  • Maniac said: Is Linode offering 512MB instance?

    No.

  • @kcaj said:
    No.

    Welp, thanks for clearing that up.

  • @Floris said:
    Could you adapt the script to support Fedora 17 or Ubuntu 14.04? I'd love to give that beast a spin. They claim it to be "up too 100 times faster"

    Floris,
    Though it took quite some time, I did adapt the script to work with Ubuntu as well.
    I tested it with multiple Vultr, DigitalOcean and RunAbove servers.

    Now the script also logs everything to a file, so it's easier to debug.
    But on Ubuntu due to some packages not being available or compile, I had to try a different route and used apt-get to install these packages ( like libswresample0)

    Right now, I tried installing it. Vultr still is on the lead. Ubuntu and Debian, they both ran the script in 26 minutes whilst DO finished in about 46 minutes.

    But, the Power8 series...
    Clearly this kind of installation-testing is not where it shines. It took hell of a time to complete. And at the last minute, it broke.

    The thing is; I realized this not so soon, Power8 series run on ppc64le architecture instead of x86_64 like all the other hosts I tried. So that itself is a huge problem.

    Why? You would ask... Well, most of the problems were due to software not being compatible with that architecture. The weren't even compiling. So I had to change a lot of parts and put some IF statements in the script. For example OpenResty didn't compile because it's Lua module is not compatible with ppc64 arch. Then I tried using nginx for ppc64 series instead. But then, ngx_pagespeed said it's not compatible either... So I had to remove that as well. Then I saw that MariaDB's official repositories only support i386 and x86_64. So I had to install that from the repo as well.

    Currently, the script "might" complete successfully in the Power8, I can't guarantee. Nor that I care that much. Because it was slooooow. Maybe if this was some test which distributes the load to all the available resources, it might've changed but... It ain't...

    About the rest, well... It works.
    With one or two small differences it works on either Ubuntu 14.04 or Debian 7.

    At least now I know how long it takes for me to build a system with Php 5.5, OpenResty (Nginx with RTMP support and a lot more), Memcached, MariaDB, FFMpeg and ImageMagick etc.
    (:

    ...
    ...

  • HsunamiHsunami Member
    edited January 2015

    Nvm, didn't read the comments above.

  • inklightinklight Member
    edited January 2015

    Have you tried Bluevm they have best records on my own tests .
    I was hoping you arrange CPU info with this result .
    As specification will different from node to node .
    Last have you tried PhpBenchmark . to see some real world performance

  • @inklight said:
    Have you tried Bluevm they have best records on my own tests .
    I was hoping you arrange CPU info with this result .
    As specification will different from node to node .
    Last have you tried PhpBenchmark . to see some real world performance

    @inklight, nope...
    I don't have any accounts with them.
    As for PhpBenchmark... I didn't even know it existed.. ^^
    But the results are so changing, even on the same server...

    ...
    ...

  • In your script you say

    We don't need apt-get dist-upgrade process to be counted

    well, you are right. But why are you counting all other "apt-get install" processes? They will alter your measurings and make them kind of useless. In fact, all other remote stuff flows into your measurements. There is no need to exclude the dist-upgrade then (add.: I wouldn't do a forced dist-upgrade, but rather check if there are any packages to be upgraded. Admin might get a real surprise when executing your script).

    TL;DR: do not execute the script for benchmarking.

  • @samguy I'm not into Debian, so can't say for sure "cloning" process that comes with this benchmark is legit. Is it?

    This benchmark also has processes that get blocked by the host node.

  • It's not a benchmark script. Let me remind. :D
    I didn't include that dist-upgrade one because some servers did took a lot of time to upgrade a lot of stuff.

    But anyway, yeah.
    You don't need to execute this for benchmarking,
    you should execute it for what it is.

    I'm contantly changing it anyway.

    ...
    ...

  • century1stopcentury1stop Member
    edited January 2015

    Nomad said: It's not a benchmark script.

    Title: A Different way of benchmarking some popular hosts...???? clone script?

Sign In or Register to comment.