5

Is there a way that we can measure performance for a host provider? If so, I would like to start this post as a wiki for everyone to chime in regarding their hosting provider. (I want to change but would like to make more of an informed decision.)

Rules:

Phase 1

  • Post comments to the Question regarding ways to measure performance.
  • Vote for comments to the Question.
  • Based on comments I'll edit the Question so that we'll agree to use a particular technique or tool to measure performance.

Phase 2

Once we've decided on the technique or tool...If you want to contribute a measurement for a host provider...

  1. If an answer/measurement for your host provider exists add your measure as a comment to that answer.
  2. If an answer/measurement for you host provider does not exist, create a new answer, list that host provider along with your measurement.

(Note: perhaps different packages for the same hosting company should be different answers...ex: Media Temple Grid Service and Media Temple Dedicated Virtual should be listed as separate Answers)

3
  • I realize that what I'm asking is kind of hard to do because how do we standardize the measurement? There are lots of variables. Are we measuring bandwidth, processing, etc. Ideally it would be great to let's say put the same page on every server and see how quickly/slowly they load. Just throwing an idea out there.
    – milesmeow
    Oct 22 '10 at 21:31
  • If we decide to use some sort of standard test page, would something like this be useful? numion.com/stopwatch
    – milesmeow
    Oct 22 '10 at 21:38
  • I think that there are probably enough people on this list for many of the hosts to be covered. I.e. if someone volunteers to host a file for download/viewing so that we can actually test the file. My friend just showed me this...linode.com/speedtest. If we come up with a sample page, then we can ask people to host it. I know that this is not the ideal test because there won't be the database component, etc., but it gets us some data for comparison.
    – milesmeow
    Oct 23 '10 at 21:23
6

Have a look at Netcraft's Hosting Provider's Network Performance, it has a lot of great information such as:

  • Downtime (in HH:MM::SS as well as a percentage)
  • Connection time
  • Time to delivery of the first byte and several other metrics.

As they've been doing this for so long (since 1995), this is probably the best first place to look for objective data.

I don't think it gives details for different products, as you asked for:

Media Temple Grid Service and Media Temple Dedicated Virtual should be listed as separate Answers

1
  • Nice link. Too bad they don't track a longer list of web hosts though. Oct 23 '10 at 11:46
4

You can measure the global response time of the providers (their home pages) with this load time testing tool: http://www.alertfox.com/Tools/LoadTime/

1
  • This is a great tool. I want to create a simple page and have people post it on their servers. Then we can use this tool to test. Otherwise, it's not a fair test...each webpage will have different content.
    – milesmeow
    Oct 25 '10 at 4:02
1

The performance of a hosting provider will depend on what you're buying from them. For ex if the bottleneck is bandwidth, that might be OK for a CPU-heavy server rendered app without many users but horrible for a popular app with a heavy js payload. If their database offering is slow, that only matters if you're using it. Etc.

The best way of producing a reliable measure would therefore be

  1. Choosing a set of representative apps (ex Wordpress, Rails, Frontend-only),
  2. Hosting these on each provider and measuring performance from different geographic locations.

How do you measure performance? This is easier because there's a lot of existing tools for measuring general website performance.

Then there are other factors like downtime, cost, availability of support, etc to consider.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.