Virtualization is the highest-impact issue changing infrastructure and operations through 2012. It will change how you manage, how and what you buy, how you deploy, how you plan and how you charge. It will also shake up licensing, pricing and component management. Infrastructure is on an inevitable shift from components that are physically integrated by vendors (for example, monolithic servers) or manually integrated by users to logically composed “fabrics” of computing, I/O and storage components (see Figure 1). This Special Report explores many facets of virtualization. Figure 1. Virtualization — The Road to Real-Time Infrastructure 
Source: Gartner (March 2008)
Server Infrastructure Virtualization As virtualization matures, the “next big thing” will be automating the composition and management of the virtualized resources. Storage has already been virtualized, but primarily within the scope of individual vendor architectures. Networking is also virtualized. The leading edge of this change is server virtualization. Roughly 90% of the server market is composed of x86 architecture servers. Based on a traditional model of one application per server, roughly 80% to 90% of the x86 computing capacity is unused at any one time. This unused capacity needs to be managed. It takes up data center space and requires power and cooling. Virtualization promises to unlock much of this underutilized capacity. IT organizations are approaching server virtualization as a cost-saving measure, and it is saving money. However, organizations that have a mature server virtualization deployment in place are leveraging virtualization for much more: faster deployments, reduced downtime, disaster recovery, variable usage accounting and usage chargeback, holistic capacity planning and more. Server platforms and related infrastructure are examined in:
Virtualization of Client Computing The key technology for defining new rules for device footprints is virtualization — a decoupling technology that breaks the close ties between hardware and software. A standard PC installation consists of a stack of multiple layers, the most important being hardware, the operating system (OS) and applications. Because of how these layers interact, the configuration of each is tightly coupled with the configuration of the layer below. This is the cause of much of the management complexity of today ‘s PCs. Because hardware changes regularly, these changes have a geometric impact on everything above. Virtualization breaks these dependencies, so the installation of each layer is independent of the configuration of the layer below. On the PC, it occurs at two levels: between hardware and the OS (machine virtualization), and between the OS and applications (application virtualization). The impact of virtualization on the PC is the decoupling of the main functional layers. Application virtualization is gaining considerable interest, because key market changes are taking place. This type of virtualization is highly valuable for dealing with current PC management challenges, but it cannot help in the personal vs. computing argument. Although more immediately accessible to you, its long-term impact will be far less significant than that of machine virtualization. This is the technology that will really make personal computing more manageable, flexible and secure by enabling users to define multiple isolated footprints on the same device. This research discusses the impact of virtualization on client computing:
Balancing Virtualization With Alternative Delivery Models IT virtualization is the abstraction of IT resources in a way that masks the physical nature and boundaries of those resources from resource users. Virtualization can take place at various points in the IT architecture, and each point of virtualization creates an opportunity for alternative delivery models — new ways of delivering capability at that layer. Virtualization technologies will make it easy to consolidate to larger resources. However, virtualization technologies will also make distributed resources easier to manage, reprovision and use efficiently. Several changes will make virtualization critical to most enterprises during the next few years. Processor capability has outpaced the performance requirements of many applications. Performance is relatively inexpensive and, therefore, the overhead of a virtualization layer is not an issue. Although processing power is inexpensive (and getting less expensive), space, power, installation, integration and administration are not, and they cost the same whether a resource is 10% or 90% used. Additionally, Web access has changed workload levels from relatively predictable to spiky, forcing enterprises to overprovision. Virtualization is not just about consolidation. By enabling alternative delivery models, new modes of providing functionality at each layer will evolve. As illustrated in Figure 2, by creating layers of abstraction, each layer can be managed relatively independently, and even owned by someone else (from streamed applications to software appliances to employee-owned PCs). Figure 2. Virtualization Enables Alternate Delivery Models 
Source: Gartner (March 2008)
The alternate delivery model aspects of virtualization are explored in:
Virtualization Affects Software Models and Licensing Virtualization technology breaks most established software pricing and licensing models. The concept of fractional use of large resources, the ability to quickly change the amount of capacity available to software, the ability to move software from one resource to another easily, and the concept of an offline snapshot (for recovery purposes, but still requiring updates and patches) become more common. Finally, there is the idea that software could be packaged and delivered in a virtual machine (VM) format, ready to run, perhaps for a short period of time. None of these new concepts fits the common paradigm of pricing based on full use of a fixed asset. Virtualized licensing presents a major stumbling block to widespread adoption of virtualization. The industry has been slow to address the problem. As vendors change their software pricing and associated license provisions to accommodate virtual use, negotiators must plan to spend an increased amount of time per contract to understand the effect of such changes on their planned software use. Clients that do not diligently monitor the ways each of their vendors is responding to virtual-use issues are likely to experience significant increased cost and the unintended impairment of their current license rights. These issues are explored more deeply in:
Operational and Management Impact of Virtualization Server virtualization challenges many of the traditional tenets of IT infrastructure, and this has ramifications on its management. First is location. Most IT resources are fixed in their placement, but virtual assets in many vendor implementations can move dynamically. Second is how they are represented — that is, as files. This makes some activities, such as provisioning, easier but may present a greater challenge in terms of security. Packaging is changing as well. VMs are being packaged as “appliances.” This removes some of the volatility that might stem from configuration variability, but the opaqueness of this approach raises issues about what is “inside” the appliance packaging. A key part of a virtual server initiative is the process of assessing server and application candidates for virtualization. The goal isn ‘t just to maximize density, but also to “right-size” workloads through an understanding of application behavior. Most enterprises attempt to place roughly similar types of workloads on the same host or server because the effects of widely varying demand could have deleterious effects on the overall quality of service. This also means it ‘s rare (but tempting) to co-mingle low-use test VMs with more-consumptive production VMs because the former environment is more dynamic, and the constant powering on and off of VMs can be felt across the entire server. It ‘s important to enact a formalized assessment process not only to predict application behavior, but also to let it run for a while in production isolation as a preventive or insurance measure. Initial placement, however, shouldn ‘t be an end to the exercise — one goal should be continuous optimization. For this, look into the VC logs (for VMware) to see when and why some VMs were relocated. Overall systems performance issues could be a sign of suboptimal VM placement. Remember that no computing activity is free in terms of its environmental costs. In addition, although the supporting virtualization technology becomes increasingly capable in terms of the types of applications it can support, conventional wisdom still says that high I/O-based environments are better left running on real hardware — for now. This research looks at the management and operational issues associated with virtualization:
Security Considerations for Virtualization Virtualization offers IT departments opportunities to reduce cost and increase agility; however, if this is done without implementing best practices for security, the resultant security incidents will increase costs and reduce agility. Security must be “baked in” from conception, not addressed later as an afterthought. The costs of implementing the best practices are significant and must be included in any analysis of the projected cost savings of virtualization. If these added security costs are avoided, the risk of not making the necessary security investments must be accepted by the decision maker in the move to virtualize. Use the security considerations and best practices outlined in this Special Report as a framework in the evaluation and selection of VM-aware security tools, and to plan, build, operate and manage secure VM environments. Existing virtualization solutions address many of the issues, but not all. It will take several years for the tools and vendors to evolve, and knowledge of the security risks must be factored into the cost-benefit discussion of virtualization. Because of the rush to adopt virtualization for server consolidation efforts, many of the issues are overlooked, best practices aren ‘t applied or, in some cases, the tools and technologies for addressing the security issues with virtualization are immature or nonexistent. As a result, through 2009, 60% of production VMs will be less secure than their physical counterparts. Research on the security issues includes: Virtualization is not simply a set of technologies buried in infrastructure. It also has important ramifications on the business use of IT, and on business itself. IT customers are often reticent about sharing equipment with other business units, and losing control of where “their” applications and “their” data reside. Ideally, customers should focus on service levels and results, not on physical location or IT methodologies. In reality, this cultural issue needs to be overcome, and can be by building success stories. The result, however, is a relationship between the business and IT that becomes more service-oriented, giving IT more flexibility on how work gets done (or how work is sourced).The ability to deploy capacity and server images virtually increases the speed of deployment roughly by a factor of 30 times. When customers are familiar with a deployment time of two months, a sudden shift to two days might require fundamental changes to business processes. Gartner clients that have virtualized heavily report that customer demand roughly doubles when speed of deployment ceases to be an issue. This increases the need for some mechanism for chargeback. The speed and flexibility of virtualization makes some form of chargeback mandatory; otherwise, demand could skyrocket and more low-priority workloads could be deployed that don ‘t justify their costs. Again, the result of virtualization will be to drive more companies to treat IT like a business, and chargeback will be a major component of that change.
|
