Doing a simple ping test to sites on Rackspace, Mosso, Slicehost, MediaTemple gs, MediaTemple dv, Pair Networks and 1and1 – who is fastest?
*Results sorted fastest to slowest… Slicehost – 48.011ms, Rackspace – 54.846ms, Pair Networks – 54.939ms, 1and1 – 57.744ms, Mosso – 60.836ms, MediaTemple dv – 93.105ms, MediaTemple gs – 93.273ms
*Make sure to read the follow up comments
These test results would be much more interesting if the testing methodology were explained in detail. What ping commands were run? Specifically, what IP addresses were they collected from? Where were they run from? What were the standard deviations of the samples? What about maximum and minimum RTT values rather than just averages. These other metrics give substantial insight into the network performance characteristics.
Also note that ping performance may be an indication of the network performance, but tells us little about how the web hosting service would actually perform on each of the different hosts.
I realize that this is not a true test of hosting performance and there are many factors to consider. I just was after a quick baseline, I ran the ping tests a couple times over a couple nights from my home Comcast connection and used the quickest response from all tests. I picked a few sites (some clients) on various hosts to test.
These are the ping commands and IP’s I used…
Slicehost – ping 209.20.80.48
Rackspace – ping 64.130.59.125
Pair Networks – ping 216.146.201.61
Mosso – ping 72.32.147.129
MediaTemple dv – ping 64.13.229.78
MediaTemple gs – ping 64.13.192.87
1and1 – ping 74.208.89.53
I repeated your test from Los Angeles using my fiber optic FiOS Verizon internet service at home, and I ended up with almost exactly the opposite result. I can conclude from this that you probably ran your ping from the midwest somewhere. The bottom line here is that this comparison is really only helpful if you’re trying to find a hosting company with a network close to where you are measuring from. It’s not a good indication of actual performance that end-users will experience, which you already mentioned by saying this is a “simple ping test”.
Here are the locations of you chosen IP addresses from your test. Note that both Rackspace and Mosso have facilities in other cities as well.
Slicehost: St Louis, MO
Rackspace: Herndon, VA
Pair: Pittsburgh, PA (?)
Mosso: Dallas, TX
Mediatemple dv: Los Angeles, CA
Mediatemple gs: Los Angeles, CA
1and1: Chesterbrook, PA (?)
The point I’m trying to make is that a simple ping test is not particularly relevant to measure the “speed” of a web host, as all it really tells you is which network is closest to you.
Adrian, thanks so much for the informative follow up – I am actually on the East Coast in Connecticut. Do you have any suggestions for the best way to test a hosts performance, that removes the originating location factor?
Dana, ran tests from my Verizon Fios (Philadelphia) and have these results (note that there is bittorrent activity on my network during these tests). Rackspace is the winner for me too.
Slicehost:
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max/stddev = 46.448/64.685/125.218/26.412 ms
Rackspace:
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max/stddev = 26.861/55.769/116.086/28.387 ms
Pair networks:
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max/stddev = 25.981/56.993/93.106/26.948 ms
Mosso:
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max/stddev = 58.822/98.881/187.639/45.372 ms
Mediatemple dv:
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max/stddev = 90.293/110.718/153.241/20.888 ms
Mediatemple gs:
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max/stddev = 94.932/121.678/174.606/28.305 ms
1and1:
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max/stddev = 54.227/88.590/131.026/27.588 ms
thanks Mike, it would be great if I could get some other testers from various locations. Anyone else care to do some ping tests?
I re-ran my hosting ping test using just-ping.com and averaged the results. Sorted fastest to slowest…
Slicehost – 137.618 ms
Mosso – 138.641 ms
Rackspace – 140.818 ms
Pair Networks – 140.036 ms
1and1 – 142.306 ms
MediaTemple dv – 151.185 ms
MediaTemple gs – 151.232 ms
Of course it’s probably best to consider where a majority of your traffic originates.
Yes, your last test is a better one because it takes perspectives from many more locations into account. That gives you an indication of basic performance of each network.
The ideal test between web hosts is something that I’ve been working on extensively. I recently used Gomez to compare the performance of Mosso Cloud Files with Amazon CloudFront CDN services.
Those results are posted here:
http://blog.mosso.com/2008/12/mosso-is-fast-%E2%80%A6-and-were-so-much-faster-its-not-even-close/
We will be doing something similar in the near future comparing actual applications hosted on various different services. So, in addition to networking performance it will be possible to see the scalability and application response characteristics of the services as well.
Keep an eye on blog.mosso.com if you are interested in that. There is a feedburner feed for it, and you are welcome to link to it from here.
It is my understanding from talking to Mosso and Slicehost that the limelight features are not always available. Slicehost is not fully integrated in the Lime Light service that Mosso is. Plus Mosso doesn’t offer RoR installs.
Pingback: 3 Easy Ways To Choose The Wrong Web Host