Computing virtualization is a popular term these days, but the concept is far from new. Back in the sixties, Star Trek's Captain Kirk had the ideal virtual computer. Aboard ship, he called out his question or command and the computer responded. He never had to tell it which drive to access, which application to load, or what routing to take when he wanted to see something up on screen. And he never worried about where or how the computer did its job.
Back when Gene Roddenberry created that show, such a computer probably seemed far-fetched. But while no one can say when someone will discover transporters, warp drives, or dilithium crystals, virtual computing is becoming a reality. The combination of the Linux operating system with clustered blade servers means that distributed virtual computing is fast becoming a means of high-performance, always available, ubiquitous computing.
A New ModelSince the inception of the modern-day computing model we have strived to create "virtualized environments" in which all resources are available in parallel to any other resource in a compute area network. In the mid-sixties the Massachusetts Institute of Technology (MIT) developed what was known as CP-67 and later evolved into IBM's Virtual Machine Architecture. It was an excellent architecture for provisioning a large set of resources centralized in a single processor. This model reduced the cost of infrastructure by centralizing administration while distributing resources.
But while the virtual machine architecture works great for mainframe computing, it did not lend itself to the complexity of services that were required for a distributed computing environment. So, for the next 40 years the search was on for a way to make the virtual paradigm available within a distributed environment. This was not just a technical issue; it was also a financial one. The key was to reduce the cost of administration, the network, storage, software, and hardware. Doing this would free up money for companies to spend on the new applications that they needed.
Now, with the advent of Linux and clustered blade servers, distributed virtual computing has become a reality. This architecture leverages resource virtualization, bringing together both the manageability of centralized infrastructure and the flexibility of compute area networks.
This concept goes by many different names. Computer Associates calls it Management On Demand, Hewlett-Packard uses the term Utility Computing, IBM calls it On Demand, and at Sun Microsystems it's Grid Computing. Whatever name you prefer to use, they all describe the ability of enterprises to deploy resources when and where needed. Rather than the old paradigm where the users send the work to a particular resource, under the new model, the resources move to where the work is.
The Technology TrioCreating such a network is not a hardware problem - enterprises already have plenty of that. They spend fortunes on idle CPUs, empty disks, and unused bandwidth just to make sure that any server or connection can handle that random peak load that might come its way. Instead virtual computing poses a management problem - how to continually reallocate resources for maximum effect. On Demand computing breaks the lock that any single device has on the ability to provide a service. Strength is then measured not by a component's specifications, but by the power of the entire network. This requires a new level of management capability.
Computer Associates, like other management software companies, has redefined the reality and economics of On Demand. Using management software on top of Linux can reduce both capital and operational costs. Also, management of Linux optimizes the availability of every application on multiple platforms. This is done by creating a complete family of software that exploits Linux clustering technology, network technology, and storage subsystems. It gives enterprises the ability to dynamically provision the compute area network.
Creating such a network requires the technology trio of storage virtualization, network virtualization, and finally, compute virtualization.
Storage VirtualizationCompanies are looking for an end-to-end storage solution that has the ability to provision within the context of the complete system - including resources external to the enterprise. That's not how it is done today. Today storage is generally provisioned on a per-user basis, without considering application usage or placement within the network. A better approach is to provision based on context rather than require a Storage Area Network or a rigid file system. This can be done using a database or metadata approach incorporating Storage Networking Industry Association (SNIA) standards for interoperability and storage management. This approach is necessary to realize such an end-to-end solution in which storage resource management extends beyond the virtualization concept and into the area of analyzing the utilization of storage across the network, the transaction, and the users.
Network VirtualizationNetwork virtualization drives storage virtualization by providing a multi-point-to-point connection that enables any system to access any storage at any point on the network. In Linux, the virtualization is realized through virtual servers using technology such as bonded dual nets and routed mesh networks. Balancing the network traffic requires a heuristic approach based on service levels. With network virtualization, administrators are able to prioritize service levels within a domain or within a system of the enterprise.
Compute VirtualizationCompute virtualization is the third component necessary to effectively achieve scaling, whereby rather than having a predetermined set of processors assigned to a particular user or task, the system dynamically assigns processors, bringing additional power to a process through the use of clustering technology. An "enterprise workload dispatcher" routes work based upon resources and utilization. This requires the ability to monitor the resource utilization of a transaction and correlate it to its service level.
These three components are the foundation for managing virtual computing.
Enter the PenguinVirtualization answers questions about utilization of the enterprise infrastructure, and the fact that infrastructure is unresponsive to changes in the environment. An underutilized and unresponsive infrastructure seriously impacts quality of service, which then impacts the bottom line. Linux helps reduce infrastructure complexity - the Linux platform delivers a common denominator that can run on multiple platforms and enables a view of Linux clusters as a single system. Virtualization with Linux drives toward having the resources from all enterprise data centers appear within a single resource pool. That pool can then be utilized in projects based on business requirements. The resources allocated to projects are determined by policy-driven services that intelligently monitor and manage the overall computing environment.
The architecture requires that the environment be broken down into provisioning, policy-driven business initiatives, automation, and use of resources directly derived from these areas. Hiding the complexity of the network, storage, processor, database, Web applications, and host-based applications is critical to creating this environment. The ability to connect all these pieces is mandatory in order to translate a unit of work within a business process to the enterprise.
The architecture enables both high availability and security. Using Linux projects such as Beowulf, Cplant, Paris, Linux High Availability, and Linux Virtual Server creates this new platform. CA is currently developing the infrastructure to manage the clustered environment, which will enable companies to use blades or enterprise servers such as Egenera Blade Frame and IBM z to leverage Linux for mission-critical applications.
The Management EvolutionA virtualized Linux environment is a breakthrough in hardware and software built for scalable applications, but it only works when overseen by enterprise-class tools offering functions such as load balancing, security, storage, and database management in a virtualized environment.
Being able to dynamically provision a user's resources and security will drive the utilization rate from the current low of 20% up to a high 90% range. The ability to dynamically shift processors to the work means a business no longer has to buy significant excess peak capacity. With dynamic provisioning of a compute area network, providing new servers can be achieved in minutes and new processors are then placed in a pool. This enables businesses to focus on the scale of applications that previously were neither economic nor plausible.
Computer Associates, along with other vendors, is providing the management infrastructure which sits above the virtualized environment and enables the evolution in enterprise-class computing. These virtualization toolsets completely aggregate processing, storage, and networking into resource pools that are allocated based upon workload. The tools encompass IP and storage networking, clustering, load balancing, hardware failover, management consoles, and secure partitioning.
Virtualization replaces static, physical hardware with dynamic software equivalents. A server is no longer dedicated to individual applications, and services are not tied to specific hardware or network paths. Virtualization software automatically reprovisions and reallocates resources based on service-levels required to meet business requirements. It provides adaptable, high-quality services at a lower total cost of ownership since it prevents resource overprovisioning.
With the management tools in place, administrators can move to a higher level of resource management. It is no longer a matter of just dealing with boxes and their interconnections. Instead the job evolves into one of strategically administering services through a geographically dispersed set of computing resources running homogeneous platforms such as Linux. And that moves us one step closer to the level of functionality enjoyed by Captain Kirk and the Starship Enterprise.
|
