Google Cloud Platform Blog: Performance testing Magento, CloudMaestro, and Google Compute Engine

Google Cloud Platform Blog

Product updates, customer stories, and tips and tricks on Google Cloud Platform

Performance testing Magento, CloudMaestro, and Google Compute Engine

Tuesday, August 18, 2015

In the world of eCommerce, website speed and efficiency matter – a lot.
Recent research1 shows that a 1 second delay in page load time equates to a 7% loss in conversions. And 81% of shoppers will actually pay more for a product in exchange for a better, more efficient shopping experience2. This level of speed and performance needs to be delivered for every customer, every time, no matter how many shoppers are visiting a site at the same time. Consistently meeting these goals requires a dynamic and responsive architecture.
At Lagrange Systems, we deliver architectures that accelerate and provide reliability for platforms including eCommerce through our software ADC solution, CloudMaestro. In order to achieve success it is imperative that the underlying cloud-based resources are as responsive as our software. When we heard about Google Compute Engine, we were intrigued and decided to set up an experiment between the leading cloud providers in an apples-to-apples comparison of relative price and performance.
The experiment went like this:

We set up and configured a simple eCommerce store utilizing the open source performance toolkit recently made available by our friends at Magento

Using their best practice guide, we defined three architectures utilizing two, four, and six application servers each, integrated with CloudMaestro

In each cloud provider, we selected similarly spec'd compute resources and collected the cost per resource.

Example Architecture with 6 application servers

Our goal: 90% of page load times under 2 seconds Our goal was to drive the maximum number of users per hour, employing the performance toolkit, that each architecture could support before the page response time became unacceptable (90% of page load times under 2 seconds). We set out to measure throughput as well as the cost to support this peak level of performance.
Even though the architectures were a fixed size, our solution is dynamic and responsive to the number of customers and the load they create on the system, so we also measured the time it took for the infrastructure to spin up new compute resources and to verify those resources are available and responding to requests.
Compute Engine delivered better performance and lower cost than the average of other tested top cloud providers Compute Engine showed up in a major way and when managed by CloudMaestro, delivered a higher level of performance AND a lower cost per user than the average (mean) of all other tested cloud providers.
Throughput Comparison (higher is better)

Cost Comparison (lower is better)

Quantifiably, Compute Engine offered up to 25% more throughput at 19% lower unit cost, meaning our customers’ dollars go further, and that Magento eCommerce sites with CloudMaestro support higher levels of traffic and more customers completing purchases.
The compute resources on Compute Engine were the quickest online, network accessible, and responding to HTTP requests. Compute Engine is over 111% faster than the next quickest cloud provider and 307% faster than the mean time of all measured cloud providers.
Comparison of Compute Resource Time to Availability (lower is better)

We were pleasantly surprised by the results and grateful to learn that whenever we need to expand our resources, we can count on Compute Engine to be there delivering what we need quickly and efficiently online.
Find out how Lagrange Systems provides reliable platforms through CloudMaestro during Cloud Platform’s webinar on 8/25, where we’ll discuss why speed and performance is key for eCommerce sites.
You can also learn more at our website, or email at info@lagrangesystems.com. - Posted by Jason Walp, Senior Sales Engineer, and Emily Friedberg, VP of Corporate Development, both at Lagrange Systems.