Using SaaS to test enterprise applications and services

Testing applications and services in the cloud really ought to be easy, for the cloud can be scaled to fit any size and performance requirement any business should require. It would indeed seem to be a business opportunity ready made for SaaS delivery – accessing such a service as and when required on a pay-per-use basis.

In practice, however, building a cloud environment that matches all the virtualisation , networking and storage requirements – particularly those required by a large on-premise datacentre running individual applications or a private cloud – has been a complex, time-consuming and above all expensive operation. In many cases, that has traditionally required the IT department to invest valuable budget in building an ad hoc test environment, or risk limiting testing to the minimum possible, coupled with a good deal of prayer.

This has prevented IT departments from being able to fully exploit the cloud for proper development and test.

Palo Alto-based Ravello Systems may, however, come up with an answer that can deliver SaaS-based applications testing of applications destined for use in large on-premise datacentres. It has come up with Cloud Application Hypervisor, which is designed to enable enterprise application developers to always test their output on replicas of on-premise production environments, and never run out of capacity.

The company claims this is the first time a tool has been available that will allow enterprises to use, seamlessly, any leading public cloud service to develop and test their existing on-premise applications.

Until now, fundamental technology differences between enterprises’ on-premise data centers and the public cloud, such as the type of virtualisation technologies used, have meant attempts at such a testing service have only been available on a technical `like-for-like’ basis. The Cloud Application Hypervisor makes any public cloud look and feel exactly like the enterprise datacentre from an application’s perspective.

It features high-performance nested virtualisation (HVX), software defined networking and storage, and an application framework. This gives enterprises the ability to create replicas of their on-premise, multi-tier VMware or KVM based applications in any public cloud without making any changes. Developers can spin up as many instances of the application as they need for testing in Amazon Web Services (AWS), Rackspace or HP Cloud, and no longer have to contemplate building out massive test capacity that sits idle most of the time.

“Most enterprises recognise the need to test on replicas of their production applications. However, it requires too much effort to recreate complex multi-tier production environments and there often isn’t enough capacity in the internal datacentre,” said Paul Burns, president and IT analyst, Neovise. “The public cloud can solve the capacity issue but it’s still a very different environment usually requiring long migration and automation projects.”

So far, over 2,000 enterprise users have taken part in the public Beta programme since it started in February this year. Between them, they have replicated more than 30,000 applications in leading public clouds representing more than 1 million CPU hours deployed. Application sizes ranged from a few VMs to over 100 VMs with multiple subnets and several virtual network appliances.

ScanCafe, the number one photo digitization and photo concierge service in the world uses Ravello to develop and test their global media management applications.

“Earlier we sometimes felt we were rolling out code a bit like we were rolling dice because we were privileging agility,” said Laurent Martin, ScanCafe’s president and CTO. “We’d rather spend our resources in developing features for our customers than in building the type of test infrastructure and automation that would be required for flawless deployments. With Ravello we no longer need to compromise; we are able to get our applications to market faster and with better quality.”

With Ravello’s application-centric, usage-based pricing model, it is now economically feasible for enterprises to always develop and test on replicas of production with no capacity constraints. Prices start as low as $0.14/hour for 2vCPU and 4GB RAM for application compute, and include the cost of the underlying public cloud.