Cluster Aware Virtualisation
Another exciting aspect of the next version is its role in the virtualization story that Novell has been unfolding during the past several months. When SUSE Linux Enterprise Server 10 came out last year, one of its most highly touted new features was its ability to run multiple self-contained virtual machines on a single physical server using Xen based virtualization.
Novell Open Enterprise Server 2 makes good on another virtualization promise that was made at the introduction of SUSE Linux Enterprise Server 10: the ability to run NetWare 6.5 as a virtualized guest operating system in the Novell Open Enterprise Server environment. As a virtualized guest operating system, NetWare now recognizes that it's running as a virtual machine. This translates into performance gains over traditional virtualized guest operating systems that require every I/O and hardware instruction to be trapped and emulated.
Virtualizing NetWare on Novell Open Enterprise Server opens up some very profitable server consolidation possibilities. On today's high-end hardware, NetWare rarely comes close to reaching full CPU utilization. When Novell Open Enterprise Server 2 comes out, you'll be able to take advantage of that under-utilized hardware by having a single machine host two, three or more of your NetWare servers without affecting performance. You can also mix and match, hosting a SUSE Linux Enterprise server and a few NetWare servers, or vice versa. As a result, you can reap one of the nicest benefits offered by virtualization - the ability to preserve access to any NetWare-dependent applications and services while you migrate your IT environment and skill sets to Linux.
Virtualization can also contribute to your disaster recovery strategies, letting you create and store virtual machine images of any supported operating system. If the physical host machine experiences a problem, those virtual machine images can quickly be loaded onto another physical host.
But perhaps the most important advantage that the virtualization in Novell Open Enterprise Server 2 provides you is the ability to maximize the performance capabilities of your high-end hardware. With SUSE Linux Enterprise 10 providing the virtualization support, you can run three different operating systems (or more if you have the server muscle) on a single physical box. That can give you significant savings on your hardware costs, rack space, cooling requirements and power requirements.
The power of virtualization is seen when you leverage its ability to run several of these self-contained virtual machines on a single compute server. (A compute server is a type of parallel processor that has no I/O except via a bus or other connection to a front-end processor that handles all I/O to disks, terminals, networks, etc.) This enables workload isolation, where instead of having multiple applications running on top of the same fat OS, you isolate each application to run on its own virtual machine. If an application happens to crash, since it's isolated, it won't affect any of the other services and applications running in their own virtual machines on the compute server. Additionally, since a virtual machine might only be running a single application or service, you will only need to load those operating system services and components that the application specifically needs. It also creates the opportunity for ISVs and integrators to develop highly customized virtual machines for the solutions they offer.
Server virtualization adds another level to High Availability automatic restart of a failed service. This means the application or service is down for a short period of time, although usually not long enough to cause production problems. Virtual Machine Migration allows an application or service running in a virtual machine to be moved from one physical machine to another in a cluster without any restart. This means there is no downtime and the complete application running state is preserved when it is moved. This is a great advantage by allowing normal maintenance of the system during production hours.
When it comes to server consolidation, the ability to run multiple virtual machines on a single compute server combined with the ability to have virtual machines running different guest operating systems can also greatly simplify those efforts. Even though SUSE Linux Enterprise Server is the host operating system for server virtualization, virtual machines can run different Para-virtualized guest operating systems. (Para-virtualization is a virtualization technique that uses a software interface to virtual machines that is similar but not identical to that of the underlying hardware.) So, for those legacy applications that need to run on top of legacy operating systems, they can be contained in their own individual virtual machine and consolidated onto a single compute server. SUSE Linux Enterprise Server can support full virtualization when coupled with forthcoming hardware technologies from both AMD and Intel.
IBM
Carol Carson, director of Linux and Grid at IBM,
There are a variety of entry points where virtualization and grid improve optimization and competitive advantage. As an organization moves through the stages, business value increases.
As you start to virtualised in the enterprise, there's a whole new level of flexibility. You now have the opportunity to have a vast array of resources available to you, without having to buy your own compute capacity. And there are a lot of software options to help with that. "Provisioning software makes sure the resources you need are going to be allocated to you," Carson said. "Orchestration software continually monitors the grid to determine if you're getting the right capacity offered to you, and the servers are being provisioned, and whether the scheduler of workload management is putting that all together for you. As you start to move outside the enterprise, there are a lot more efficiencies to be gained, as well. By being able to share data with partners and suppliers so you're not doing this as two separate operations, you can speed your time to market.
"When you look at the characteristics of grid evolution, there are a number of key tools that are important as you start to progress. On the homogeneous front, good cluster scheduling is really key, as well as having good cluster file systems, file and storage virtualization. As you start to move toward heterogeneous environments, advanced scheduling becomes even more important. Now you're dealing with multiple applications, multiple resources that are all vying for a piece of the pie.
"It's important to have integrated security and provisioning. You need tools that determine you need X amount of servers and to provision those servers with the scheduler and the orchestration software to predict capacity requirements.
"Workload management -- to be able to move work to where it can most efficiently be run. And information virtualization -- one of the most important things as you move along is to get the new business insights from the formerly disparate pieces of data."
Virtualize like resources -- Homogeneous systems, storage, and networks
Virtualize unlike resources -- Heterogeneous systems, storage, and networks; application-based grids
Virtualize the enterprise -- Enterprise-wide grids, information insight, and global fabrics
Virtualize outside the enterprise -- Suppliers, partners, customers and external resources
"When you move beyond your immediate department," Carson said, "things like meta-scheduling become extremely important. Most folks grew up with individual silos. For example, you might be using Data Synapse, somebody else might be using Platform, and you need to have things communicate.
That's why, as we start moving toward virtualised heterogeneous environments, standards become so important.