|
http://www.dmst.aueb.gr/dds/pubs/jrnl/2005-IEEESW-TotT/html/v29n5.html This is an HTML rendering of a working paper draft that led to a publication. The publication should always be cited in preference to this draft using the following reference:
|
The virtual machine (VM) is the most dazzling comeback in information technology. IBM implemented a VM platform architecture in the late 1960s in its CP/CMS operating system. The company’s goal was to provide the time-sharing capabilities that its batch-oriented System/360 lacked. Thus a simple control program (CP) created a VM environment where multiple instances of the single-user CMS operating system could run in parallel. Thirty years later, virtualization was rediscovered when companies like VMware found ways to virtualize the less accommodating Intel x86 processor architecture. The popularity of Intel’s platform and the huge amount of software running on it made virtualization an attractive proposition, spawning within a decade tens of proprietary and open source virtualization platforms.
Virtualization has progressed a lot from its primitive beginnings. Today, virtualization lets us run most modern operating systems in a VM that can be hosted on facilities ranging from our laptop to a datacenter in the cloud. Once an operating system runs as a guest in a VM host (also known as hypervisor), it becomes easy to control via high-level operations. You can save its image into a file, move it from one host to another, launch multiple clones, suspend it until it’s needed, share it with others, rent it as a service, or ship it to a customer. Techniques such as paravirtualization, in which the guest offloads some heavy lifting to the hypervisor, allow the guest systems to run efficiently and share hardware with a minimal waste of resources. Virtualization means never having to beg for a server.
As developers, we can benefit from virtualization technology in several ways: better developer environments, more efficient testing, easier distribution, and improved infrastructure management.
Increasingly deep and wide application stacks have made provisioning a development environment a tricky task. Modern applications often depend on many layers of large infrastructure chunks that might include the application development framework, third-party libraries, a relational database, a message broker, scripting languages, the revision control system, debuggers and tracing tools, and logging facilities. Each component can introduce other dependencies and may require careful installation and configuration, while successive versions of the application might depend on diverse specific versions of infrastructure.
With VM technology, it’s possible to create a virtualized development environment that contains all the tools, applications, and libraries that a programmer requires. A standardized development environment packaged and distributed as a VM can help bring developers up to steam in minutes rather than days. All that programmers have to do is download the development environment matching the product version they’re working on and fire it up on their desktop workstation or even their laptop. Tools like SUSE Studio and the more specialized Vagrant simplify the specification and creation of such a development environment.
Apart from convenience to developers and operations staff, a virtualized development environment offers many other advantages:
System designers and administrators also benefit from the availability of many prepackaged third-party VM images. These allow the painless experimentation with a complex setup: just download the VM image, fire it up, and you’re ready to go. Popular virtualization platforms come with tens to hundreds of VM images of pre-installed operating systems, development environments, or end-user applications. By using these, you can defer the investment for properly installing and configuring a system in your organization to the time when you’ve actually settled on the one you’ll use.
Virtualization is also a boon when testing a system. Ensuring a pristine testing environment is tricky, especially after you’ve run many tests and haphazardly modified your setup to replicate a specific error. With a VM image of the testing environment, things become easy. You start each test or testing session on a fresh machine instance booted from the testing VM image. When you finish testing, you simply discard the instance, never to see it again. Handily, when you nail down a difficult to replicate bug, you can save the VM image where you performed the feat to pass it on to a developer for further examination. This technique and the parity between development, testing, and production systems eradicates many of the “works on my machine” reasons for not fixing bugs.
In addition, the ability of testers to launch on their computer VMs corresponding to diverse platforms, makes it easier to test your product’s portability and compatibility. In the past, such testing would require maintaining a testing laboratory with physical instances of each machine.
VM images also simplify the deployment of your application. You can package all your system’s components and setup into a so-called VM appliance, which your users can download and run at the touch of a button. Many cloud infrastructure providers support the hosting of such images, further simplifying the distribution. This practice eliminates problems that stem from the interaction of your application with the customer’s setup and minimizes the support required to tailor your system to a customer’s environment.
However, deploying applications as virtual appliances is not without problems. Once you go that route, you become responsible for keeping up to date all your system’s infrastructure included in the VM: the operating system, third-party libraries, and other applications. If that requires deploying a completely new VM, you need to plan for how to transfer your customer’s content and customizations from the old VM image to the new one. This might not be too difficult for stateless systems such as firewalls, spam filters, and caching proxies, especially if you provide a managed solution for storing their settings on your own infrastructure. On the other hand, patching together applications that run on separate VMs is more complex than if they were running on the same instance. On balance, deploying applications as virtual appliances is an attractive proposition for delivering stand-alone demos—but it’s not a panacea for production setups.
Virtualization tremendously benefits the staff handling your operations, but this doesn’t mean that you as a developer can’t pick up some winnings in this area as well. Simplified and more effective infrastructure management means less work to support such features within your system. Virtualization can easily provide many features that older systems once provided through heaps of native code. Application provisioning for higher-throughput could just mean firing up more VM images of your system behind a load balancer. System maintenance under high availability requirements can often be performed by running the new VM side-by-side with the old one on the same hardware and adjusting the load balancer settings. Disaster recovery is also simpler when you can regularly backup a VM’s image to a remote site. Finally, the utility of virtualized instances is compounded by the fact that they can be run in the cloud through services such as the Amazon Elastic Compute Cloud and the Rackspace Cloud Servers. This means that you can set up your organization’s infrastructure without tying up capital and resources on procuring and maintaining physical assets. But this is a topic for another column.