Taking .NET Development to the Cloud
The leap of faith
Jun. 17, 2009 02:45 PM
Cloud computing platforms, such as Microsoft Azure, offer compelling advantages for building new scalable .NET applications. But can the Cloud be used for developing existing .NET applications? In this article, I'll explain how we've made the leap to Cloud-based development for our internal applications and the lessons we've learned along the way. Specifically, I'll describe our checklist for selecting a Cloud vendor and how we've used the virtualization capabilities of the Cloud to improve our agile development process. I'll also outline the quantifiable benefits we've seen, including saving $100,000 in capital expenditure and reducing our iteration cycle times by 25%.
All Clouds Are Not the Same
As the development team lead for Buildingi, a corporate real estate consultancy that specializes in back-office technology solutions to manage large portfolios, I'm responsible for building Web-based applications using Visual Studio, the Microsoft .NET Framework, and Silverlight. Last year we started looking at Cloud Computing to gain the advantages of a scalable, virtualized platform for software development and testing.
Most of our developers were using virtual machines (VMs) on their individual development workstations, which limited the ability to share environments across the team. In addition, giving access to customers for user acceptance testing was still a very manual process that involved taking new builds and installing them on under-utilized physical servers.
We had ruled out an onsite virtual lab solution because we estimated even a modest lab would cost well over $100,000 in hardware, storage, lab management software, and labor, which was well beyond our budget of $40,000. By moving to the cloud for a new development and test lab we hoped to avoid this large capital expense and gain the ability to:
- Quickly set up and tear down multi-machine environments
- Snapshot entire systems and store them in a central library to enable more efficient bug replication by our regionally distributed development team
- Run multiple isolated copies of the same application stack side-by-side for testing
- Enable customers to access new software builds over the Web
Microsoft Azure is a Platform-as-a-Service (similar to Google App Engine) and won't support existing multi-tier .NET applications unchanged, so we explored using Amazon EC2. EC2 is classified as Infrastructure-as-a-Service and offers Xen-based virtual machines and Cloud-based services, including storage, database, and queuing services. Essentially, this means you get Xen-based virtual machines in the cloud, can scale these machines on-demand, and only pay for what you use.
We loved the potential of EC2, but the fact that it only offers Windows 2003 and no other Windows clients (such as XP and Vista) was a non-starter for our client/server solutions. We found the user interface very limited and not suitable for use by an end customer. It also lacked workflow functionality to easily share entire application development stacks with customers or across the team.
Another requirement to consider was the ability to import VMware images of an application stack to a Cloud and run them unchanged. EC2 will only import its AMI format and this is only possible for Linux-based images. After some more research, we discovered Skytap's Virtual Lab, an emerging start-up that offers a virtual lab as a Cloud-based service and solved these requirements.
A Virtual Lab in the Cloud
Virtual lab management is a term coined to describe a new set of tools and test practices to automate labs using virtualization technology. There are vendors (such as VMware) that offer virtual lab management as an onsite package and Skytap offers these capabilities in the Cloud. Virtual labs offer some important capabilities over individual workstation virtualization solutions:
Virtual machine images are the containers that enable operating systems and applications to be isolated from physical resources. A group of virtual images that defines a complete system, including network and storage characteristics, is called a configuration. For instance, a configuration could consist of multiple Vista client machines, an Oracle database server, and a .NET application server. A configuration is a very useful concept for QA teams because it allows a whole system to be defined and isolated in a test environment. Virtual networking enables copies of the same environment to be run in parallel and the emulation of production environments during the test process. Many Cloud vendors only support single machines, so finding a solution that supports configurations is important.
A configuration library allows a team to manage and organize virtual images and configurations. Standard builds and images can be created and made available to development and QA teams to save manual installation and set-up time. Additionally, the library is used to store new configurations that are cloned or created as part of a test.
Suspend, Snapshot, and Restore
The ability to suspend a complete state of a multi-machine configuration and make a snapshot (a copy at a point in time) is a major benefit of Cloud environments. This is especially useful for application development teams because when a bug is found a configuration snapshot to be taken at the point of failure and a link to the configuration added to a defect report. Instead of a developer spending hours to reproduce the defect, he or she can restore the configuration and start debugging the issue in minutes.
Automating a test lab almost always involves integrating tools and test processes. An automation API enables teams to automatically create test environments as part of the build process and initiate automated test runs once a new build has been deployed. An automation API is typically made available through a Web Services interface.
Administration, Reporting, and Security
Administration and security features include user and quota management, project creation, permissions, and authentication. Remote access to the system (for instance, for an outsourced vendor) is usually managed through secure connections via encrypted protocols and virtual private networking. Reporting lets users and administrators manage usage and quotas to keep project costs in budget.