Some interesting figures have emerged from the Gomez Division of Compuware, which demonstrate just how pervasive the cloud is becoming, even for those users who think they are not having anything to do with the cloud at all. The figures also demonstrate that, without effective service testing, company brands and reputation could stand at risk.
The unseen cause of potential problems is that while a customer business may assume its cloud services provider is the host for the contracted services, that is actually rarely the case. The information content in many services can come from a wide variety of sources.
For example, according to Imad Mouline, Chief Technology Officer of the web performance division of Compuware, nearly 20% of applications currently widely used in business already use content that is hosted by Amazon’s EC2 service.
“We looked at some 40,000 applications and the percentage currently using at least one piece of content hosted on Amazon had risen from 16.8% a month ago to 19.6% yesterday,” he said. And when it comes to just Amazon-hosted content and services, Europe is right up there as source. “Amazon’s US-based East Coast facility is the most common source for content, but second is the company’s Dublin facility. This is followed by the US West coast and then Asia,” he said.
“What we test are the most critical business processes and support codes for business,” he said. “We test them over and over again, from 30 locations worldwide, 24 x 7, to ensure it performs well – and if not tell them why.”
Gomez can differentiate which part of the process – applications, servers or network - are not performing well. This is based on synthetic testing – a robot that is performing the same test to a high level of consistency, with the tests performed from two different locations. One will be robots running in large datacentres with the largest systems resources and the best network connections and fully calibrated. So if there is a problem in the application delivery chain it can be identified. The other will be a smaller, less resourced location
It can also test the `last mile’ around the globe. Gomez has robots running in end user PCs in 125,000 locations worldwide, in 160 countries. These are consumer grade PCs and laptops which are not optimized in any way. Gomez effectively rents resources to run tests in background mode on end user PCs and when that system runs a test, Gomez pays the owner a fee. So customers can select operational areas to test – for example a US company could test the efficacy of an online marketing campaign in Europe.
“We can do this to a fine level of granularity, so it could be as granular as testing a marketing campaign in Paris. This means users can use the service before going into production to see if the service is viable, or after going into production to check the service is there and working as planned.”One area where the company exploits its end user network is in real world load testing of an application or service. Test PCs are selected on the basis of location in order to map onto the required target markets for the application or service. It is then subjected to a series of tests launched from those PCs, giving what is, in effect, a bi-directional result of both the end user experience (in terms of response and page servicing times) and a real world load testing of the service under pressure.
“And when it does start breaking down we can tell why, where and what is causing the problem,” he said.
Typical cloud problems include service providers running out of resources, and it could be the involvement of third party service hosts involved in the delivery process. At the time of talking the average was 8.86 in the US. According to Mouline, in Germany that number is up to 10, with each one being an individual host service contributing to the delivery process and sending data direct to the browser.
“This could be just 100 bytes of data,” he said, “but if it does not get loaded the user does not get their page.”
This issue, he suggests, is about to get worse with the increasing use of mobile devices. The next generation of smart personal clients are going to ramp up both the number of applications and services delivered to those platforms and, as a consequence, the bandwidth required. In addition, the nature of mobile phone service offerings is likely to lead to a much greater number of contributing hosts being involved. This inevitably means that service degradation and temporary loss are likely to move from annoying possibility to infuriating probability.
And though the root causes of such service degradations and losses will be some contributing service provider somewhere, the impact will be felt by the business that has its brand on the service as delivered to the end user. So it is in the interests of both the brand owner and their primary service provider to ensure that they have, between them, a good understanding of the service they deliver, its actual service delivery record and its points of weakness.
Gomez provides this type of information for clients as a paid-for service, but is aware that many businesses won’t need the level of testing and reporting that it can provide, at least not to begin with. But there will be a need to get a basic idea of what is happening with the delivery of any service so that some idea of what is normal can be established, and some idea of the nature of problems identified when they do occur.
To service this requirement, the company recently launched its free CloudSleuth service. This is available in beta form to anyone online at www.cloudsleuth.com and provides a toolset capable of providing a business with basic monitoring of cloud-based applications and services. It also gives a real-time picture of the current response times of the major public cloud service providers, such as multiple Amazon locations, OpSource, Rackspace GoGrid and Microsoft Azure, around the world.
Post new Comment