8 Comments

Summary:

A new visual dashboard by Cedexis provides a glimpse into what clouds and content delivery networks are performing best at a given point in time. Cedexis’ aims to help content owners pick the best infrastructure for their cloud workload on an ongoing basis.

ispScreen Shot 2012-04-16 at 12.45.59 PM

A new visual dashboard launched by Cedexis on Monday provides a country-by-country, vendor-by-vendor view of CDN and cloud performance.  A tool like this is important as companies evaluate moving more workloads to the cloud or clouds because it gives them some basis for comparison.

Update: Here’s a taste of Cedexis’ dashboard showing cloud provider stats:

Cedexis was founded in 2009 by two Akamai veterans to help businesses choose the best clouds and CDNs for their needs on a dynamic basis. If there’s an outage or slowdown on one cloud, the Cedexis service can move that workload elsewhere. One customer saw 40 percent page-load improvement just by using Cedexis for network optimization — and this was a sophisticated customer that was already using advanced cloud and CDN technologies, said Cedexis co-founder Julien Coulon.

For the report, Cedexis collects data from a few hundred customers that embed a piece of JavaScript code on their websites. Cedexis aggregates data  “about the performance, errors, throughputs from every network in the world, putting it out in a way to help them choose which [cloud or CDN] they should work with,” Coulon said via phone from the company’s Paris office. “This is real data from real end users who need to see what’s happening. And we’re doing this transparently in a way that helps the cloud and CDN providers improve their quality of service.”

“No provider is good everywhere, all the time, for all users, but if you mix them right you can get significant improvement in your web site performance,” Coulon said.

Cedexis sells its service to large content providers, including French media companies Canal+ and FranceTelevisions and luxury consumer products makers Dior and Piaget. U.S.-based customers include Mozilla. On the monitoring side, Cedexis competes with Compuware’s Gomez  and Keynote Systems. On the network load-balancing side, it vies with companies like F5 Networks.

Amazon is the cloud giant that most companies evaluate when they want to make a move, but the company doesn’t provide a ton of insight into its inner workings.  While some might think that Amazon and Akamai, the CDN leader, might take umbrage to  statistics that might show them in a less-than-stellar light, tools like this could also help them to improve their services. Amazon, which suffered a serious outage last year which Cedexis profiled here, might also welcome the opportunity to show more about its performance, going forward, as more competition from the OpenStack players comes online.

Update: And with that, here’s another taste of the Cedexis data, this time on CDNs:

You’re subscribed! If you like, you can update your settings

  1. Hi Barb, the images shown with your post lists ISPs, not CDNs or cloud providers. Akamai, Level 3, Limelight, EdgeCast, Amazon etc. are all missing from the charts. Maybe Cedexis is tracking them in another chart, but what is being shown in the images are all ISPs. I can’t find any chart by Cedexis that shows CDN providers.

    1. you’re right — it was operator error, in my haste i picked the wrong charts. story is updated w/ charts on Cloud providers and cdns. thanks for your note!

  2. @Dan : Here is the good link wit all informations http://www.cedexis.com/country-reports/

    1. Thanks Niko. Where can I find the methodology on how the testing amongst the CDNs was done. What kind of file was being tested? Many of the CDNs listed don’t do streaming so is testing being done with small files, large downloads, HTTP traffic? And in what regions of the world? And over what periods of time? All is says is “response time”, but of what? And something could respond fast, but not deliver the content with good performance, so are you testing the whole network, to the last mile, or only the backbone?

      The charts look nice, but without all of the methodolgy behind them, we really don’t have any idea how valid the data in this chart is.

  3. Many of the CDNs listed don’t do streaming so is testing being done with small files, large downloads, HTTP traffic? And in what regions of the world? And over what periods of time? All is says is “response time”, but of what? And something could respond fast, but not deliver the content with good performance, so are you testing the whole network, to the last mile, or only the backbone? Posted by – http://www.usajobsnew.com/

  4. Great Questions! Cedexis Radar tests from the perspective of millions of web visitors all over the world. http://www.cedexis.com/products/radar.html

    Technically, we first deployed 2 test objects to every major CDN and Cloud Player, one small (about 50 bytes) and the other “large” (100KB). You can see them here:
    http://www.cedexis.com/cdx10b.js
    http://www.cedexis.com/cdx10b-100KB.js

    We have over 200 customers who have deployed the Radar tag. When someone visits their sites, the browser instantiates the Radar client. The Radar client waits until 2 seconds after the onLoad() event and then goes and downloads some instructions from us. These instructions include our complete list of benchmarks plus any private platforms specific to that customer.
    http://www.cedexis.com/products/measurements.html

    The Radar Client chooses 4 of the benchmarks and performs 3 measurements.

    1: HTTP Connect
    First Radar measure how long it takes todownload the small object. This measurement includes local DNS resolver cache hit or miss, if we are measuring a CDN then also the caching node cache hit or miss plus TCP socket set up time. Some of these are not in the control of the provider being measured so we tend to see more variability in this measurement.

    2: HTTP Response
    As soon as that first download completes we download the small object again, reusing the open TCP connection and the browser’s cache for DNS resolution. This is HTTP Response time and we consider it to be a fairly accurate representation of basic latency from the browser to the provider.

    3: HTTP Throughput
    We download the 100KB object, again reusing the TCP connection and browser DNS cache. We measure Kbps based on start and finish of the download over time.

    At each step in the chain, we “cache bust” the local bowser cache by appending a random number to an HTTP GET query string.

    Lastly, we calculate an Error Rate based on which of these downloads succeed or fail.

    1. Based on Cedexis data, Cedexis CDN is very interesting. Have you seen any independent/external test of their CDN ?

  5. One thing I’d be more interested in besides these averages is something that somehow showed consistency….maybe something on the order of standard deviations. I’ve used a number of CDNs, and having gone through all the cheap ones which often times can be faster than Akamai, they all suffered from fairly regular ‘hiccups’ where maybe one connection out of 20 would hang for long periods of time. Akamai ruled in terms of consistency…

Comments have been disabled for this post