The nightmare scenario for system administrators working on web services and applications: they push a new iteration of their code into production, everything seems to be going well, then, as demand peaks, their cloud servers slow to a crawl, users start complaining, databases can’t keep up with the demand.

The code push introduced a regression that wasn’t caught in testing and under real world use, the service goes down. Perhaps it was a caching flaw, an error with the way files were pushed out to content distribution networks, or a coding mistake that caused the application to use far more resources that it should.

Just a few years ago, that was a risk that every web service provider took. There was simply no way to run tests accurate enough to account for every edge case and every load scenario. Sure, they could throw a few test cases at their development or staging infrastructure, but it was impossible to create a test load that accurately mimicked real world scenarios with multiple users connecting from many different servers all across the globe.

Federated Clouds Make Load Testing Easy

Times have changed. Federated cloud platforms provide the perfect testing infrastructure. They allow companies to build massive but ephemeral testing networks that can accurately replicate real world loads.

Federated clouds can be thought of as clouds of clouds. An integrated network that connects together the platforms of multiple cloud providers, allowing users to spin up virtual machines in any location around the world, run their test workloads, and then spin down the servers.

Before cloud integration platforms like ComputeNext were available to businesses, a testing infrastructure like this would have been both hugely expensive and hugely complex: the expense would often outweigh the potential benefits. Managing the interfaces, APIs, and billing portals of very different cloud providers, often in different countries, was not a task that many businesses were willing to take on.

Cloud integration layers that allow for the straightforward and intuitive deployment of federated clouds are paradigm changing for stress and load testing scenarios. A single interface and API allows for the fast construction and dismantling of geographically distributed virtual machine networks that used for scripted load testing, accurately representing the sort of loads that can be expected when code goes live.

Even better, because ComputeNext also integrates billing, there is a single point of payment and costs are much more predictable.

ComputeNext provides internet service and application providers with the tools to accurately and extensively load and stress balance their code, putting an end to the “push and hope” scenarios that have been the cause of sleepless nights to many a system administrator.

 Image: Flikr/myoplayer