Hosting benchmark, bogoMips

When you are doing a hosting benchmark there are thousand common parameters all providers suggest to compete against others (size, bandwidth, #databases… ).

The main secret of the providers is the speed. All providers hide this information: CPU speed and number of people hosted on the same server. These 2 parameters can vary, so at the end of the day the important parameter is the real time response.

I found the concept “bogowips”, which is a play on the Linux “BogoMips”. It is short for “Bogus WordPress Instructions per Second”. This concept was created by a WordPress developer that has invented his own criteria to measure it. The utility he has created is this plug-in:

http://wordpress.org/plugins/benchmark/

BogoMips comes from Bogo Millions of Instructions Per Second. The results in my case are:

  • CPU Speed: 24.686 BogoWips
  • Network Transfer Speed 4,24 Mbps
  • Database Queries per Second: 1.049

I love the idea but for me it still is not so much credible because I do not know where the average of the other data comes from (it could be the case that they come from dedicated VPS). In any case it is a good idea that I hope that evolves in something more consistent.

I still hope there will be a standard speed measure recognized as standard that enables the community to evaluate the time response of the services they hire.

Who are your IP neighbors?

Time response continues causing me headaches, but I’m discovering causes and differentiating the causes, so I’m progressing in the right direction.

The situation now is that pictures are compressed and with a light weight, no code faults, reduced CSS and JS files, CDN activated, cache, etc.

It depends on the pictures of the main screen that time response differs, anyway the time to load the page yesterday was 3,26s. What’s the main issue?

The first 2 requests take 1,3s (40%) mainly the web browser waiting for response from server (yellow line).

joapen-time-responseSo, the server, what’s happening with the server?

After checking for some weeks the performance of the server, the conclusion is that memory assignment is fine, and there are not faults on the server, the parameter that provokes time response issues is the CPU at 100%.

What does it mean 100%? I have not the data of the speed and number of CPUs, but I have read about strategies of hosting companies about how they co-locate the shared environments and how they do it.

Basically, you can host so many people in one server and be aggressive on price, or host less people on same place and increase the price (providing better quality). I have read about hosting services that co-locate 5-10 people on a server, and other that co-locate over 400 places.

You can easily understand how much neighbors (places) you have in your neighborhood (the shared server) using tools such this:

http://www.dnsqueries.com/es/vecinos_servidor_web.php

I work with two different hosting providers, in one I have 29 neighbors, in the other one just 4.

Now I know:

  • I have to move to the neighborhood with less neighbors.
  • Next time I look for a hosting service, I will ask for the number of neighbors.
  • The consumers community of shared hosting services should start demanding a clear parameter to understand the quality of the shared hosting. Some offers include infinite disk or bandwidth, what about a minimum CPU assignment in MHz?