Server Virtualization and the Hybrid Cloud
By now, we think you understand what virtualization is all about. You know that virtualization is so important because it decouples the software from the hardware. Decoupling means that software is put in a separate container so that it is isolated from the operating system. Although the cloud infrastructure doesn’t require virtualization, virtualization is often a component of a cloud deployment. The point is that virtualization is important in the cloud because it provides a level of abstraction. The abstraction doesn’t need to be virtualization, per se; it might be done another way (like a grid), but you need a way to do this abstraction in order to provide functions for end users.
Clouds are often laid on top of this virtual server substrate. It is a foundational technology. When building a cloud infrastructure (whether you’re building a private cloud or a public cloud), you generally think about it as a distributed environment that’s dynamically allocated or provisioned. Virtualization enables this distributed environment. Virtualization is most concrete on a server level. For example, this is really what IaaS is all about — a distributed resource that can be dynamically allocated.
Whether to virtualize your infrastructure depends on the workload you’re dealing with. For example, some experts caution that if your workload is running analytics across thousands of nodes at full capacity on a big data problem, using a virtualized environment might not make sense.
As companies build out their virtual data centers, many find that the next logical move is to implement a private cloud. This provides them with more flexibility around provisioning resources. However, in a hybrid model, your company may want to bridge your private cloud to a public cloud. Here are some examples:
Application bursting: When spikes in resources for a particular application are required that overwhelm your private data center, you may want to borrow resources from the public cloud.
Moving virtual machines: It’s exciting to think that you might be able to balance loads between private and public clouds at the virtual machine level. In this way, you could run them where they’re the most effective.
Ideally, in a hybrid cloud there needs to be a shared, standardized environment or at least some kind of consistent virtualization infrastructure between your on-premises private cloud and what exists in the public cloud. Several approaches have emerged:
Compatibility: Large vendors such as VMware, IBM, Microsoft, and HP are putting offerings on the market that enable enterprises (and service providers) to move virtual machines between on- and off-premises infrastructures. Often, these infrastructures use the same technology platforms as you have on-premises. This makes it easier to be consistent. VMware is even certifying cloud providers that offer a globally consistent IaaS service that can be provisioned via a GUI interface.
Virtual Private Cloud (VPC) services: Cloud vendors like Amazon are offering a VPC service. The Amazon VPC, for example, lets you provision a private, isolated section of the Amazon Web Services (AWS) cloud where you can launch AWS resources in a virtual network that you define.
Open source software: These open source projects are looking to bring together public and private cloud infrastructures. One example of this is Eucalyptus (www.eucalyptus.com
). Eucalyptus utilizes virtualization software to create resource pools that can be scaled up or down. A Eucalyptus private cloud is fully compatible with Amazon APIs, which makes it easier for organizations to migrate virtual servers between a private cloud and a public cloud environment.
Cloud management services: Other third-party providers, such as RightScale (www.rightscale.com
), are providing cloud management services. For example, RightScale provides a cloud configuration framework that are templates to configure and operate servers. It also provides an automation engine to help provision and monitor server deployments.
Desktop virtualization
Just as virtualization refers to separating resources and services from the underlying physical delivery environment, conceptually, desktop virtualization refers to separating the desktop environment from the physical machine. This kind of virtualization has gotten a lot of attention over the past few years.
The term desktop virtualization can be somewhat misleading, given that organizations are making use of all kinds of devices — smartphones, tablets, notebooks, and PCs. In fact, desktop virtualization is sometimes called client virtualization or even endpoint virtualization. Also, the desktop virtualization model is changing. Whereas several years ago, people talked about virtualizing a desktop to a server, and this is still a primary desktop virtualization model, newer models are emerging.
Virtualizing the desktop can happen in a number of ways. On the server side this can include
Terminal server: In this approach, a desktop is provided to a user from a remote server. The operating system is not stored on the device. Only the screen image is actually transmitted to the user, who may have a thin client (it has no hard disk) or possibly an old PC.
Operating system streaming: In this approach, the OS software is passed to the client device — but only as much of the software as needed at any point in time. Technically, this process is called streaming. Some of the processing occurs on the disk and some in local memory. The Windows OS, for example, and its applications are split between the client and the server. Streaming applications run at about the same speed as does reading the application from the disk. You can use this approach by utilizing PCs on a desktop or by using thin clients. The model requires a constant connection, which is why it’s typically used with desktops.
Server-hosted virtual desktop: Here virtual PCs (complete emulations of a PC) are created on a server. The user has what appears on the server to be a complete PC. In actuality, the server-hosted virtual desktop is remotely executing the operating system on the server, and the graphics are being sent to a desktop over a network. Sometimes people refer to this kind of client virtualization as a virtual desktop infrastructure (VDI). This is one of the most popular forms of desktop virtualization. VDI provides the capability to have shared client sessions on the server rather than on the client. The software you need to use sits on the server, and an image can be viewed on your device.
For example, in healthcare, clinicians are using VDI desktops to gain access to information in patient rooms and offices. In science labs, where space is at a premium and contaminant-free work areas are a priority, virtualized desktops eliminate the server and other hardware from the room. Other examples include using virtualized desktops for temporary workers or remote workers who need access to applications, and even for traders who need to move around the trading floor but need access to information on demand. Moving the desktop to the data center (whether a public or private one) covers every possible means of replacing physical PCs with graphics terminals (also called thin clients).