Originally published on March 07, 2011 by Dirk Paessler
Last updated on January 23, 2024
•
5 minute read
One of our test and demo websites for PRTG is a 5-way cluster which has cluster nodes in California, Texas, Ireland, Singapore, and Australia. We use this cluster to monitor a selected set of cloud servers and CDN services for uptime and request times. See my blog post from January for more technical details about the setup.
In January we found that CacheFly CDN and Google AppEngine followed by CloudFront, Rackspace Cloudfiles and SoftlayerCDN were the best performers. Did these 5 continue to outperform the others? Let us look at the data from the last 30 days!
Since my last post we have added Google Storage for Developers (Beta) as the 21st contender (see here for a complete list).
Request Times
My first graph shows the average request times of the services and servers in milliseconds. In 28 days we have made one request per minute (one request from each one of our globally distributed nodes every 5 minutes), so these results are based on almost 1 million requests in total. 
In a global perspective the three fastest are Azure CDN, Cotendo CDN and GoGrid, all are - at an average - below 200ms around the world. The next segment are Google Storage, Google AppEngine, CachelFly and Highwinds CDN which are all below 500ms. 
Downtimes
Let's look at the most reliable services. The following graph shows the aggregate downtime of the services as observed by our cluster: 
The best result comes from Amazon S3 EU: zero downtime in 30 days. Then there are three segments:
- Below 30 minutes we have Amazon S3 EU, Google AppEngine, Amazon S3 US, Rackspace CDN, Cotendo CDN, EC2 US East, Flexiscale Cloud, Joyent Cloud, GoGrid/Edgecast, VoxCast CDN, EC2 EU
- Up to 100 minutes: CacheFly CDN, MaxCDN/NetDNA, SoftLayer/Internap CDN, Azure CDN, EC2 US West, Amazon Cloudfront, Google Storage (Beta), Highwinds CDN
- The bottom segment (above 100 minutes) are Level 3 CDN and Cloudflare CDN (more than 6 hours of accumulated downtime)
Combining Downtimes and Request Times
This final graph shows all services in a two-dimensional chart. We would want a service to be close to the bottom left where the fastest request times are combined with the best uptime: 
The Verdict of March 2011
In March the top services are Cotendo, GoGrid, Google AppEngine and Voxcast. Last month top performer Cachefly was set back by 51 minutes of downtime. Without being offered as a CDN Google's AppEngine is again one of the top performers.
An honory mention goes to Amazon S3 EU for truely 100% uptime.
You there!
