In honor of Cloud Connect I’ve taken some Cedexis Radar data and made suggestions for those considering Azure and Amazon as cloud vendors. How good are their peering arrangements? How effective are they at delivering dynamic content to the last mile?

When you try to ask this question using Cedexis Radar, the first thing you realize is to be meaningful, you actually have to ask a more granular question: from which ISP and over what time frame? The per ISP and per minute variability in Radar measurements are a constant source of amazement to me. So to start, we’ll choose the week of Monday, January 30 2012 and 3 major US ISPs: AT&T, Verizon and QWest. We’ll measure HTTP Response Time – a measurement we define as the time to request and download, over a warmed up socket, a 50 Byte file. I’ve included both of Azure’s US locations, all 3 of Amazon’s US locations and for comparison the Internap/Voxel US location.

In each graph, the Y axis is average HTTP Response time in milliseconds measured from millions of browsers sitting in each network.

We’ll start (of course) with AT&T, the largest US network by subscriber base:

AT&T US (ASN 7132)

Interesting that EC2 US-West starts well and trails off and that Azure NorthCentral outperforms EC2 US-East. I’d have guessed that the AWS Virginia facility would be the consistent winner.

Verizon US (ASN 19262)

To be clear, this is Verizon’s home and business network (not their mobile network). Surprising that, except for a midweek hiccup, Voxel’s US location edges out Amazon US-East.

Qwest US (ASN 209)

In this case EC2 US-West is consistently the fastest even though that US AWS region would have been a poor choice for reaching visitors coming from AT&T or Verizon (at least over the course of this week) and Voxel’s US location is as much as 40% slower than the top choice.

Of course, our stance is spreading your risk and your performance across multiple cloud providers is the best bet and we’ve build a set of tools to make multi-provider strategy and execution possible. We can even show you how well your own peering providers are doing. Wish I’d had this in the last job!

This integration opens up the possibility of using any measure that New Relic offers, such as memory, CPU, database, and Apdex score to add a new level of intelligence to Openmix applications. For example, imagine a scenario in which a data center is reaching a critical load threshold. Openmix can be programmed to direct fewer requests to that data center before the data center starts exhibiting performance problems, based on the information supplied by New Relic.

For the demonstration we crafted a simple Openmix application that chose randomly between two data centers unless one of the data centers reported CPU load above a certain threshold. If the threshold was exceeded then that data center was removed as a possible destination for traffic. In this chart you can see the behavior:

• From 23:15 to 23:27 “New Relic Provider 2″ was decided against because it was exceeding the CPU threshold set in the Openmix application
• From 23:27 to 23:48 both providers were chosen (at random)
• From 23:49 to 00:08 only “New Relic Provider 1″ was used because the other provider was over the CPU threshold

You’ll note that the chart is noisy and that the number of decisions is very low. The noise is because of the randomness of selection when both providers are candidates, and the low number of decisions is due to using dig only once per second during the demonstration.

We’re excited about the potential for this integration and others like it.

You can schedule a meeting with him now : jason_at_cedexis.com

However, with great power comes great responsibility so we need a testing tool capable of harnessing and double checking each application. Simply writing an Openmix Application and loading it into the portal presents a number of insurmountable testing problems. Imagine you have written an application that weights or removes Providers depending on the client country of origin? Imagine you are uploading capacity data into your app via Pulse Load and you want to balance in such a way as to avoid bottlenecks. Once the app is loaded and you are pointing live traffic to it, how do you know if your business logic is being applied correctly?

We recommend PHPUnit. With PHPUnit, you can emulate the live data feeds and check that your app responds in the desired way. For example, you can write a test like: “when the network of origin is in China and my Singapore datacenter is more than 500 ms in HTTP Response time and ChinaCache is faster than CDNetworks I expect Openmix to choose ChinaCache based on my rules. You can construct as many tests as you like to ensure your business logic is performing correctly. The cool thing is you can do this all from your laptop, prior to rolling out to production.

PHPUnit was written by and for some pretty hard core PHP developers. To make getting started easier head over to the Openmix Application Library and download the openmix-sdk  and the openmix_library.

Now point your browser to the SDK and follow the instructions for installing PHP and PHPUnit . If you unzipped the SDK on your Desktop, your browser URL will look something like this on a MAC:

file:///Users/christopherhaag/Desktop/openmix-sdk/openmix-sdk/doc/php/api.html
Then unzip the openmix_library.

The best first step may be to cd into one of the example apps and trying running PHPUnit. For example, if you have downloaded the library to your Desktop, the steps to follow are:

1. $ cd Desktop
2. $ unzip openmix_library.zip
3. $ cd openmix_library
4. $ ls -F openmix_library  # here are the 5 first examples, each with a working unit test
   avoid-by-geo/        conditional-host-name/    geo-with-overides/    perf-and-avail/        perf-with-penalty/
5. $ cd per-and-avail
6. $ ls -F
   README    api/    app/    test/s
1. # note the 3 directories
2. app/ # contains the actual Openmix application
3. test/ # contains the PHPUnit test application and a helper script
4. api/ # contains the wrapper code for PHPUnit which allows it to emulate the live Openmix
5. # now run one of the tests
7. [chaag@mac-air:perf-and-avail]$ phpunit –verbose tests/OpenmixAppTest.php
   PHPUnit 3.6.6 by Sebastian Bergmann…
   Test: 0Test: 1Test: 2Test: 3Test: 4Test: 5Test: 6

   Test: 7

   Test: 8

   Time: 1 second, Memory: 5.50Mb

   OK (2 tests, 28 assertions)

Congratulations! You’ve run your first PHPUnit test and all 8 tests have passed! Now return to the SDK and explore the extensive docs and then try your hand and modifying one of the examples apps  and example unit tests

In the future, we’ll be migrating this all to GITHUB and putting a discussion portal in front of each app. Happy scripting!

Cedexis history was made yesterday, when we passed the 1 billion mark in daily recorded Public Radar measurements. Over one billion (1,060,620,742 to be exact) pieces of public website performance data were recorded for Jan 11th, 2012.

Why is this number important? Well, perhaps the number itself is not, but the fact that we are growing this daily tally, is! With each piece of Radar data, we are improving the collective knowledge of our customer’s end-user online experience. Over the last two years, Cedexis Radar customers have been adding to the ever-growing set of data that is in turn used by our Openmix platform to help shape their Cloud, CDN and Datacenter traffic patterns. The larger the number of Radar measurements, the better our understanding of web site users experience is, the better traffic shaping service we can provide.

Here’s to 1 Billion daily points of data, and counting!

This kind of incident happens when an ad campaign is launched, a link is published on website with a strong audience (thus the “Slashdot” effect) or when a TV show is watched by millions of viewers at the same time.

The case recently hit a Cedexis customer from the luxury industry. The case showed that a large part of the viewing audience came from the Internet.  While millions of viewers watch TV, an increasing number of view content from their laptop, tablet or smartphone.

Whilst the brand did not wish to be named publicly, its E-business CTO made an important choice to support the sudden burst of online traffic. His secret ?  A delivery strategy which could handle the traffic surge thanks to a Cedexis solution which can guarantee optimal performance.

You recently appeared in a prime time TV Show, how did you prepare the event ?

Customer : We didn’t prepare anything in particular. We are in the most critical period of the year and we prepare our platforms to sustain high-traffic loads.

Did your media teams alert you in time to let you prepare any particular mechanism in case of high traffic boost ?

Customer : No, not this time. As an international group, we are not always informed of all broadcastings, and this is the reason why we have to be ready at all times. Our architecture is redundant and tailored to absorb sudden high traffic peaks depending on marketing campaigns (displays, emailing, etc.), viral effects or TV broadcasts. In order to achieve this, we rely on cache application systems as well as on various CDNs upfront. These are not only a user’s response addition solution but they can also absorb traffic bursts much better than what platforms and/or web hosting partners can offer. However, we had previously noticed that each CDN has its own problems and cannot be always good all the time. That’s why we need to ensure an optimal user experience at all time. In order to achieve this intelligent and real time load balancing, user’s requests are dispatched towards the best source thanks to the Cedexis decision platform.

How did your web site perform concretely during the TV show ?

Customer : We experienced a strong audience from the beginning of the broadcast, visitors increasing by 7 fold, page views by 5. The overall activity started to decrease at the end of the show and it took a little more than 2 hours to get back to regular traffic loads.

Did you notice any slowdown in performance ?

Customer : We didn’t notice any slowdown from a user’s point of view. The monitoring probes didn’t detect any noticeable deterioration. However, our bandwidth went six times higher with a peak level at 130 megabytes per second. Our server’s activity also showed some increase but the majority of our contents was cached and the servers were thus able to sustain the load. Precisely, Cedexis helped us to refine our cache strategy during this critical phase for our brand. We were incidentally able to route user’s requests to our only CDNs, thus reducing the load on our origin host and on our applications architecture. The total of available data, including the ones brought by Cedexis Radar and OpenMix, helped us to control the load balancing, viewers level peaks and response times experienced by our users.

The impact to your customers and business will depend upon 3 factors:  your cloud provider (in this case, Amazon),  your application, and your cloud orchestration layer.

The first factor, your cloud provider, is largely out of your control although you could certainly exercise your right to share strong opinions with your provider.  In all likelihood, the smart folks at Amazon will listen to their customers, as they’re known to do well.

The second factor, your application, can be written to tolerate some level of disruption in service.  A resilient application could be clustered in locations (within the same cloud or across multiple clouds).   During this scheduled maintenance, an application could redirect requests to another cloud location and restore state from an unaffected region.  Amazon provides guidance on writing failure resilient applications.

In case a week is not enough time for you to rewrite your application or convince Amazon to postpone the reboots, you could consider the role of your cloud orchestrator.  A service like Cedexis Openmix shields web users from the impact of cloud reboots.  Here’s a recent example of how the impact Edgecast network availability was averted using a standard dual-provider strategy with Cedexis in the middle.

Are the Amazon reboots making you consider mutli-cloud options?  If so, contact Cedexis today (while you still have time). We’re the experts in global multi-platform cloud strategies.  We’d love to hear from you and give you the visibility you need to keep your service running and your users coming back.

$1 for the first TB — 98% savings!

According to the pricing page, the regular price in US and Europe would be $61.44 for the first TB of traffic But wait, there’s more!  (Cedexis data, of course.)

Taking a snapshot of Cedexis Radar public probe data from today, we see that Net DNA sits just outside the middle of the pack on network latency @ 283 ms response time.  Happy holidays from @cedexis and find yourself some bargains for your content delivery too!

From these views, a few interesting points emerge.  Amazon’s brand is in the forefront for AWS, while Microsoft takes a backseat to Windows.  While “service”, “elastic”, and “AWS” occupy a predominant place in the Amazon message, “applications”, “cloud”, and “sign” (and “service”) take the forefront for Azure.

Here at Cedexis, we’re tracking the public clouds, private cloud (data centers) and CDNs for our clients the world over.  Come help yourself to some free data and answer critical cloud migration questions for your enterprise.

To achieve this goal, TVTrip had to revamp its architecture originally based in OVH’s French data center. Traffic growth and hotel’s video presentations require a very specific attention. Technical teams must help to provide a flawless online experience to users and to B2B partners who are renting content, and use it on their own website, engineered by a proprietary technology of TVTrip.

Considering this, the CTO first went with a CDN in order to increase performances while leverage the load on the infrastructure. But many issues arised, and amongst all the mandatory supervision of the quality of user’s experience, as well as taking over the web traffic control in case of service break down or slow performance delivery from the provider.

Challenge

At first, TVTrip could count on independant monitoring from its providers using a probe (agent). The information and data were insuffcient however to monitor the true customer’s experience.

« We detected a real issue in the lack of detailed information delivered by some CDNs », explains Charles Gourlaouen, CTO at TVTrip. « We compared the available products and signed for a solution that offers a virtual experience of our service from the user’s point of view, using remote probes all over the world, but this solution came out to be expensive and we didn’t get the whole picture », adds Charles Gourlaouen.

TVTrip then decided to look for a service that offers a true user’s experience view. He found Cedexis to hold his hand on this issue by integrating in the begining of 2011 its community-based monitoring & analysis tool Cedexis Radar.

« After a few weeks, the results were speaking for themselves », the CTO explains, finally satisfied by the measure of performance from the user’s view but also the performance of the providers (web host, cloud and CDN). Getting the information is all good, but to use it for a real time issue is even better, «  that said, taking to consideration we deliver the equivalent of 70 million videos a month through our site TVTrip,com and our distribution network », said Charles Gourlaouen.

Usage of Cedexis

After using data from the Radar, TVTrip wished to increase its Cedexis collaboration by experiencing the OpenMix decision taking platform, to mix various providers usage and try out multi-sourcing.

TVTrip content (HTML and CSS) is generated and essentially delivered by the origin (OVH, a server provider in France, virtualized servers from Amazon). The millions of images and videos are managed by Akamai and Amazon.

In order to guarantee a top level quality of service and to monitor drops in performance, Cedexis took the performance data to route each user to the most performing provider thanks to the OpenMix DNS-based Global Traffic Management/Global Server Load Balancing (GTM/GSLB) platform.

« We use OpenMix to balance the load between providers, taking performance into consideration depending on the type of traffic », spots Charles Gourlaouen.

Content delivery depending on the origin of the request was one of the key points to take into consideration. Joint usage of Radar and OpenMix shows here too all its capacity to adapt to specific needs : « The nearby factor is important for us : we can now, for a partner in Asia using our contents, control traffic to route users to the best local provider being the closest possible to the providers », TVTrip’s CTO delights himself.

For the less popular videos, Cedexis offers the possibility to route traffic to a more cost efficient provider and sometimes a more optimized one, for a specific audience (of a specific geographic origin).

Results

The launch of Cedexis service suite has been beneficial on numerous key points (source TVTrip) :

• Providing visibility on the real user’s experience and visibility on the quality of service of the providers/CDN : Cedexis becomes a trusted partner.
• Flexibility like no other over web traffic control : possibility to better deliver content following specific rules, depending on the type of customer and the type of content.
• Help with decision taking and valuable advices : Cedexis brought to TVTrip an independant consulting on its technical architecture and possible future evolutions.
• Help with providers/CDN negociations : usage of Cedexis is a real leverage to costs optimization and to prevent single-vendor dependancies.

« Cedexis certification in itself helps to have a broader discussion with CDNs and acts like a true insurance to guarantee the best performances at optimized costs », rejoyces Charles Gourlaouen.

His conclusion is that « thanks to this approach, we can faster detect the origin of an issue using actionable data, thereof can we tell our customers what the problem is. It was important to be able to prove that the origin of a problem was not coming from TVTrip and then protect ourselves from issues that could have impacted our B2B customers. We are confident that our image and trust are not broken and as a result we are not losing contracts. »