Despite its many advances, the modern data center is still a complicated mess of technology silos and components that are manually cobbled together and managed. This complexity imposes tremendous operational burden and cost. Fortunately, data centers are going through a major transformation, driven in large part by virtualization. This sea change promises to simplify and automate management, allowing IT to focus less on data center plumbing and more on delivering IT services that drive the business forward.
Today's Data Center versus Yesterday's Manufacturing Plant
This revolution is not unlike major innovations in other industries. Take automobiles and their manufacturing plants as an example.
Yesterday's "made to order" manufacturing processes were manual, labor-intensive and time-consuming. Henry Ford revolutionized automobile manufacturing by standardizing and automating the production line. Subsequent innovations such as "just in time" inventory and lean Six Sigma processes have further improved efficiency. These innovations made it possible to manufacture automobiles quickly while improving quality and the end customer's ownership experience. Today, highly sophisticated cars built to individual customer specifications churn out of automated production lines with amazing precision and speed.
Automobiles themselves have transformed due in part to advances in manufacturing and adjacent technologies. They are more reliable and feature-laden than a Model T owner could have imagined (and owners had plenty of time to imagine while manually cranking the engine). They run for hundreds of miles without refueling and thousands of miles without maintenance. Consoles notify drivers of the car's needs and potential hazards while sitting comfortably in an air-conditioned cabin. Much of the automobile itself is self-managing, from auto-adjusting headlights to traction control and all-wheel drive when needed.
Today's IT "factories" or data centers and the IT services they produce are in need of dramatic change. Data centers operate well below capacity and are dependent on a maze of complex underlying infrastructure and an army of skilled personnel doing manual, repetitive labor. Most IT services are custom built with little standardization or efficiency, and keeping them running requires a lot of manual cranking just like the first automobiles. The result is a tangled mess of brittle components and a lack of responsiveness to the business that increasingly depends on IT.
Today's Data Centers: Component-Level Management
Inefficiency spans the entire lifecycle of an IT service, from the initial business request to retirement. End-user requests set in motion a long, multistep development process that can takes weeks or months. IT waits for equipment to be delivered. Once it arrives, a manual and time-consuming process begins to assemble the many disparate hardware and software piece parts. Hardware sizing is uncertain, usually resulting in oversizing. Hardware resources are difficult to redeploy once committed to a specific application, so re-use and re-configuration are inefficient and unlikely. As Jim Jones, a network administrator at a telephone company in Kansas, puts it: "You'd have to size the machine, plan on it being bigger, buy it, get it in, and it's not like the last machine you got, so you can't deploy your standard image on it. You have to install your operating systems from scratch, get your drivers rounded up, mount the servers in the racks and possibly disrupt service if you start knocking cables around. Then you have to make sure you have enough power."
Once an IT service is running, there are similar challenges of disruption and inefficient processes that create risk. Administrators set alerts at various infrastructure levels, but it's usually the end user who discovers a service-level problem in the form of slow or spotty performance, and calls the help desk. The help desk routes a trouble ticket for investigation. Administrators troubleshoot the problem, often by physically going to the user's desktop, and attempt to fix it. Understanding the causes of performance degradation caused by the infrastructure, as opposed to the desktop itself, is difficult and time-consuming and impacts service levels when not resolved. Any testing or troubleshooting is typically a manual, error-prone process.
As this description illustrates, IT is very good at seeing the trees, well-practiced at measuring and managing piece parts and components. As with today's automobiles and their manufacturing plants, tomorrow's data centers require a fundamentally different and better approach that manages the forest, not the trees. Virtualization is driving that change, raising IT management above the level of components to the level of the services that drive and, in the end, build the business.
Tomorrow's Data Centers: Virtualized and Automated IT Services
Virtualization has fundamental advantages that create an opportunity to rethink and reinvent management. To start, virtualization abstracts all the physical (x86) infrastructure including servers, storage, and networking so IT services (applications) are not entangled with the underlying components (aka complexity). This new software layer treats these underlying resources as a common pool of resources available to any new IT service. This standardization and abstraction makes setting up new services dramatically simpler. Virtualization also delivers many of the requirements of IT services such as high availability and security-these are imbedded in this common layer and can be attributed to applications with a mouse click (remember, this is simply software).
Virtualization can change the game for IT management because it can more effectively manage what the business cares about-complete IT services-at a higher level than the individual piece parts. This new model for management allows business owners to easily request new services and specify desired service levels (availability, security, and performance) while enabling IT to deliver on those requirements quickly without custom building each IT service.
Similar to today's automobile with its traction control and air conditioning, tomorrow's virtualized IT services will be much more self-managing. IT will be able to automate, regulate, and monitor service level compliance through policies that are built into the IT service definitions provided by software vendors or end users. Under this approach, a transparent charging model will allow the business to know up front what service will cost and give IT the ability to report regularly on the cost, quality, and consumption of IT services.
A Unique Approach to Service-Level Management
Virtualization brings a new level of simplicity to IT with a service-level management model that enables end users to request services and enables IT to manage complete IT services more holistically. This journey will bring new levels of simplicity and automation to computing. In the future, IT will act more like a service provider with the ability to seamlessly interact with both business owners (who request new services) and end users (who consume them), clearly understand and manage their expectations, ensure that infrastructure delivers the services efficiently without a lot of manual intervention, and transparently allocate costs for the services delivered.
While this sounds like a great marketing story that has been told for years, what is new are the inherent advantages of virtualization which can be used to make it possible. Making this an operational reality, however, requires the ability to 1) encapsulate an entire IT service into a standardized software container, 2) describe service levels required by the IT service and 3) dynamically manage service levels through policy-based automation. Initial capabilities and key building blocks in each of these areas are available today.
Virtualization platforms and management offerings deliver built-in automation that dramatically simplify IT management, streamlining or eliminating many tedious day-to-day operational tasks. What once took weeks or months now takes minutes. Infrastructure can be provisioned and service levels managed with unprecedented automation. Security and availability can be delivered at the click of a button as an attribute of the infrastructure. IT can respond and add value to the business with greater speed and reliability. This simplification has increased administrator productivity for customers, improving their server to administrator ratios, and freeing staff to pursue new strategic initiatives (or go home at night and on weekends!). Yet there is still a lot of opportunity to further automate and reduce complexity at all layers of the technology stack and for different roles in the data center.
Managing in a virtualized world will also make it much easier for customers to seamlessly take advantage of external or public clouds without outsourcing IT. Because virtualization encapsulates entire IT services and service-level requirements can travel with those services, customers will be able to move to an external cloud environment in an evolutionary way without rewriting applications or ripping and replacing today's infrastructure. Portions of today's internal clouds built on virtualization can be migrated to external clouds to take advantage of needed capacity or better economics, without being held captive by the underlying technology.
Companies are moving from a component-level management approach to a service-level management model, and they are doing so by first transforming their data centers with virtualization. While the promise of service-level management is not new, virtualization can truly deliver it with its inherent advantages of abstraction, encapsulation, and standardization. The challenge is to harness the power of these attributes and raise the level of management to entire IT services and the service levels end customers and the business require of them. As a result, IT will function as more of a "utility" service provider and can evolve nondisruptively to cloud computing.
The transformation of the data center that is underway and driven by virtualization is a journey that has unfolded previously in other industries. Manufacturing plants and automobiles are but a few examples. Thankfully, the journey will take us to a much better place, where IT departments and IT services are not held hostage by the complexity of managing the underlying components. This brighter and not-so-distant future is something to keep in mind midway through assembling the piece parts of a new IT service late at night or driving down the freeway in a sophisticated and air-conditioned automobile!
Rob Smoot is responsible for marketing vCenter data center virtualization management products at VMware.