Virtualization is a paradigm-changing technology that is going to affect how IT infrastructure will be managed to ensure cost-effective delivery of end-user services. In simple terms, virtualization is an abstraction that decouples IT services from their underlying computing, storage and networking hardware dependencies. This abstraction enables a logical presentation of resources that considers hardware as a single pool of processing, storage and networking power, simplifying management for IT administrators.

Virtualization offers flexibility and fluidity to seamlessly scale and manage the infrastructure based on varying demands and at the same time maintain the desired service quality levels. It allows heterogeneous resources to run side-by-side on the same physical machine while maintaining isolation between virtual machines. These resources can then be managed remotely and optimized globally by administrators and managers across the enterprise - lowering total cost of ownership (TCO) and increasing efficiency - while maintaining a seamless, high-quality user experience. Virtualization promises significant benefits to businesses and adoption is on the rise. Yet how do IT organizations take advantage of this technology for wide-scale production and successfully manage it on an ongoing basis?

Rise of virtualization in the enterprise

Server virtualization projects commonly fall into three categories - hardware virtualization or hardware emulation; OS virtualization or OS partitioning; and para-virtualization. Hardware emulation, characterized by VMWare and Microsoft Virtual Server, is a host-based approach that runs multiple OSs from a single server, effectively partitioning a single server into multiple machines. The benefit of the hardware virtualization approach is the ability to add extra server capacity without additional hardware purchases, run multiple applications on the same physical host server, and easily move virtual machines from one physical host to another.

OS virtualization, exemplified by companies such as SWSoft and Sun, takes a base OS and runs multiple instances of that OS on the base. The major benefit of the OS virtualization approach is that all OSs share the same hardware resources; it is also a leaner architecture and a good method to consider for those IT organizations consolidating or deploying multiple virtual servers on a single Linux, Windows or Sun physical server.

Finally, para-virtualization is characterized by vendor XenSource, an open source product that is similar to hardware virtualization but runs "guest" kernels.

Para-virtualization offers similar benefits as hardware virtualization, but customizes the OS for lower overhead and increased efficiency.

Server virtualization is certain to sustain and build momentum in 2006 because of the multiple benefits it delivers to the IT organization. A recent survey by IDC, for example, predicted that more than 75 percent of all large companies (larger than 500 employees) are in the process of deploying virtual servers, and current users of server-virtualization technologies report that 45 percent of new server purchases in 2006 are expected to be virtualized. Virtualization enhances IT management efficiencies and reduces maintenance costs by enabling the consolidation, flexibility and security of deployed applications and infrastructure leading to better alignment of IT resources with business needs. For example, when used for testing and development set-ups, virtual machines allow a new level of safety and security as developers using multiple OSs and platforms can easily "reset" a virtual server if an OS is corrupted, and set up a virtual OS if needed, within minutes. The ability to improve performance and resource utilization, speed up provisioning of new services, rapidly develop and deploy applications while avoiding downtime due to outages or maintenance windows are all technical benefits that ultimately drive the bottom line, especially for revenue-generating, crucial enterprise applications.

Server virtualization also provides significant benefits for disaster recovery and business continuity considerations. Studies show that the cost of downtime to organizations can run anywhere from $50K per hour/day to $6.5 million per hour/day depending on the industry, and this doesn't even take into consideration the potential longterm effects on customer satisfaction and business reputation. Virtualization is a cost-effective alternative to setting up and maintaining duplicate systems, especially for small and medium sized businesses with limited resources to devote to business continuity plans. Because virtual machines emulate the complete environment including data, applications and OSs, without affecting the hardware, outages in one area of the network do not affect service delivery - presenting continuous availability to the user, even when outages occur.

Finally, many enterprises are considering server virtualization as an overall network virtualization strategy, in which major areas of the enterprise IT ecosystem including storage and applications are virtualized as part of an overall initiative to improve control, efficiency and flexibility. This overall virtualization strategy is being adopted quickly in the marketplace. A recent IDC estimate, for example, puts virtualization activities at nearly $15 billion worldwide by 2009. Server virtualization provides not only a near-term benefit, but in the long term can pave the way for enterprise network virtualization, better service quality management (SQM) and application performance management (APM), linking the entire infrastructure.

In order to reap these benefits, managing server virtualization should be considered carefully and comprehensively. Server performance management - real-time and historical workload monitoring, capacity management and trending and service level reporting - is a critical component in virtualization initiatives. The next two sections review the major initiatives within the virtualization process, the key points that must be addressed in order to be successful and the beneficial role that service-centric performance management can play in ensuring that a virtualization initiative is successful and yields the anticipated results.

From Pilot to Production

Driven by the compelling value proposition, a number of IT organizations have started pilot implementations of virtualization technology in an attempt to realize higher infrastructure ROI. Virtualization is still an emerging technology and the industry has not yet reached a level of maturity, with deep experience in the system, processes and tools, to tackle the challenges of managing wide-scale virtual infrastructures. As a result, some organizations are hesitant to move from pilot to the production stage because of a lack of technical expertise and process maturity, inadequate integration of virtual infrastructure management with standard operating processes, poor visibility and inconsistent performance monitoring practices. Virtualization represents an organizational and operational paradigm shift where IT administration is required to deal with near real-time changes in the infrastructure driven by application demands and powered by the fluidity of virtualized platform. IT administration, accustomed to managing the more static "physical" infrastructure, can perceive the dynamic nature of virtualization technology as a source of uncertainty.

To date, the challenge of managing virtualized infrastructure has been to cost-effectively meet application service level objectives (SLOs) in an environment where an increasing number of business-critical applications are competing for the shared infrastructure. A service-centric performance management solution can tackle a number of these challenges to help facilitate this transition from pilot to production stage. The key service-centric capabilities include: Performance Management, Automation, Capacity Planning, Predictive Analysis and Change Management. These capabilities can be applied across the entire management lifecycle to realize the full potential of virtualization.

Performance Management

The ability to make virtualization decisions and ongoing workload optimization decisions based on performance data is key to a successful virtual infrastructure deployment. In addition, it is essential to monitor the server infrastructure in the context of application performance to ensure its alignment with application requirements.

Performance management is necessary for the entire infrastructure, including physical servers, virtual machines (VMs), host, and host farms. Monitoring performance at the host farm level is essential to make decisions related to reallocation of VMs across hosts within the same host farm. Monitoring the Key Performance Indicators (KPI) at all levels along with built-in analytics is also required to accelerate the problem resolution process. Monitoring performance at the application level is required to ensure delivery of desired service quality to the end-user, analyze the potential impact of planned infrastructure changes, and quickly resolve any issue in case of performance degradation.

Automation

With virtualization, new applications can now be provisioned in seconds rather than days; change request response times are cut down to minutes; hardware maintenance can be accomplished with zero downtime without waiting for a maintenance window. Such a drastic change in operational requirements demands increased automation.

A number of virtual infrastructure management decisions currently based on an administrator's dexterity and practical experience can now be automated using a policy-based approach applied to service-centric performance data. Service-centric performance data is an essential element in the automation of virtual infrastructure management tasks such as dynamic provisioning and archiving.

Capacity Planning

Without conducting a proper capacity planning exercise, the risk to the business can be quite high. It is essential to right size the infrastructure and have a plan to match the IT forecast with the changing application demands. Inadequate capacity increases the risk of violating SLOs, while excess capacity increases the total cost of infrastructure and reduces overall efficiency.

When virtualization is applied, server utilization rates are increased from 15-20 percent in a typical environment to easily more than 60 percent. However, by using a service-centric performance management solution with powerful capacity planning and predictive analysis capacities, utilization rates can be increased even further and maintained at around 80-90 percent.

The key enabler for effective capacity planning is baselining and trending. The performance baselines and trend analyses form the basis for decisions for architecture changes, physical to virtual machine transitions, dynamic resource reallocations and virtual machine migrations. Such decisions require historical and real-time intelligence on information such as usage, resource performance and application performance.

Predictive Analysis

The increase in flexibility and fluidity of infrastructure offered by virtualization, demands powerful predictive analysis capabilities. This allows IT administrators to predict potential service performance degradations and fix the issues before they impact the users. With dynamic provisioning of resources, infrastructure can be incrementally changed in a timely fashion in order to maintain the cost-effectiveness of service delivery.

Change Management

Virtualization represents an environment where there is constant change. It is essential to not only maintain an accurate list of all configuration items (CI) including VMs, hosts and physical servers, but also to track and accurately record changes to any of the CIs. This includes changes to server (physical or virtual) configuration, host configuration, movement of VMs, etc.

A service-centric performance management solution with powerful discovery mechanisms offers the ability to discover various CIs, host farm clusters in a datacenter, and linkages between guests and hosts, to allow the IT manager to have aggregated views of hosts and linkages to virtual machines.

A service-centric performance approach

One of the important requirements for a successful server virtualization project is to be able to seamlessly manage the entire server infrastructure, physical or virtual, supporting the full lifecycle management. The operational and management lifecycle for a traditional infrastructure consists of four stages: Readiness Assessment, Planning and Design, Implementation and Ongoing Workload Optimization and Support.

With the use of the right tools and processes the challenges inherent in each stage of the lifecycle can also be addressed in a virtual infrastructure implementation to lead to success. Whether dealing with physical or virtual infrastructure, a number of decisions across the entire management lifecycle are driven by application performance and infrastructure ROI. A service-centric performance management solution that supports both virtual and physical servers is a key asset in each of these four stages.

Stage 1: Readiness Assessment

The key objective of this stage is to assess if the server infrastructure is ripe for virtualization and if the benefits can be fully realized. Achieving this objective involves analyzing the existing infrastructure, reviewing the overall health of the IT assets, identifying which applications and servers are candidates for virtualization and developing the roadmap that outlines the transition of physical servers to a virtual infrastructure.

The key consideration that drives such decisions is server capacity utilization and performance, looking at not only the current snapshot but also the historical trend. In addition, it is important to assess the application demands on the servers and the degree of fluctuation of these demands, in order to ascertain how well they can leverage the fluidity of a virtualized platform to ensure that SLO commitments are met. This stage also involves identifying the risks, issues and barriers associated with moving to a virtualized environment.

In this stage, a service-centric performance management solution can help in the following ways:

analyze existing networking infrastructure leveraging comprehensive real-time and historical performance reporting along with powerful analytics,

identify candidate servers ready for virtualization utilizing pre-defined workflows, and

assess risks associated with moving to a virtual environment using predictive capabilities

Stage 2: Planning and Design

The planning and design stage involves architecting a scalable solution and then creating a detailed implementation plan. The key objective of this stage is to align business demand to IT strategy.

It involves choosing an optimal VM configuration to transform a physical server to a virtual platform, sizing the hardware resource required to host selected virtual machines and clustering the hosts into a host farm. It is important to ensure that infrastructure performance and service quality performance as perceived by end-users is not negatively affected by virtualization.

In this stage, a service-centric performance management solution can help align the virtualization strategy with the changes in business demands in the following ways:

Physical to Virtual transition analysis: Conducting a scenario comparison to achieve the optimal configuration and combination of VMs on a given host machine, and multiple applications stacked on the same hardware.

Impact Analysis: Determining if the application performance will be affected by the underlying virtual infrastructure performance.

Baseline and Trend Analysis: Tracking application performance trends, determining which VM or hosts are most and least utilized, and determining when and what to upgrade based on historical capacity trends.

Stage 3: Implementation

The implementation of a virtualized environment includes conducting a pilot consolidation, validating the design, deploying the hardware as planned, configuring VMs and migrating data to the new system. In addition, it requires deploying configuration and change management systems to accurately discover and track all CIs.

During the implementation stage, IT must employ monitoring capabilities to ensure application performance and server performance are not negatively impacted as a result of virtualization. Any design deficiency, performance issue or hardware problem detected during this phase must be quickly resolved.

In this stage, a service-centric performance management solution can help ensure successful implementation in the following ways:

Implementation of discovery and change control mechanisms: Automated discovery and change management mechanisms can help maintain an accurate list of all CIs including VMs, hosts and physical servers, and accurately track and record changes to any of the CIs.

Before-and-After Analysis: It is essential to understand how the applications and infrastructure perform before and after virtualization. Through a combination of availability management, impact analysis, and before-and-after analysis, the IT organization can assess performance to ensure that the application and infrastructure performance is not degraded as a result of virtualization.

Stage 4: Ongoing Workload Optimization and Support

Service-centric workload consolidation and optimization is an integral part of ongoing support to fully realize virtualization benefits and maximize ROI. The key objectives of this stage include cost-effective delivery of agreed service quality levels, and the ability to ensure high availability and accelerate problem resolution.

In order to achieve this, IT organizations must have real-time control of VMs, be able to optimize host resource usage for all guests, and determine which guest to move from one host to another and when. In addition, IT must be able to evaluate the health of both virtualized and non-virtualized infrastructure and IT assets, and constantly ascertain and evaluate the problems within the entire virtual infrastructure, ensuring that if degradation happens, IT can quantify the impact through impact analysis and move quickly to problem resolution.

In this stage, a service-centric performance management solution can help ensure successful implementation in the following ways:

Performance data driven Workload Management: Real-time and historical performance data can help IT administrators objectively make the right optimization decisions in terms of workload management. Some of the workload optimization tasks such as provisioning of resources on demand can be automated using a policy-based approached based on historical performance data. The tool can also develop intelligence as it increasingly correlates between performance data trends and workload management steps.

Predictive Analysis: Pre-emptive troubleshooting is an essential ongoing capability for high quality service delivery. Predictive analysis of the server, VM or host enables proactive problem resolution and pre-emptive troubleshooting so that performance and service degradations are detected and fixed before they affect the end-user. Threshold monitoring and alerts based on historical and trend data support this capability.

Problem Resolution and Impact Analysis: If degradation happens it is essential that IT can identify the problem, quantify the impact through impact analysis and help prioritize the problem resolution efforts based on the impact at the application and business priorities. An effective performance management solution will isolate the problem, offer dynamic troubleshooting drilldowns and correlate application, VM and host performance for effective understanding of the root cause. In addition, it is essential to have ongoing monitoring and assessment of the network health through regular monitoring of predictive indicators. The ability to effectively isolate and troubleshoot problems helps avoid SLA violations which affect the end user experience, as well as regulatory or financial penalties that may result from these violations.

Conclusion

An optimal performance management solution will guide IT managers through the lifecycle for server virtualization deployments, a key first step to a larger virtualization strategy for the enterprise. With the right planning and implementation capabilities, as well as ongoing support, any organization can quickly realize the substantial benefits associated with virtualizing their IT infrastructure.