Virtualization Feature Story

But OS-level virtualization stops short of delivering a complete virtual infrastructure. OS-level virtualization is great for cutting up a server into multiple small partitions. Unfortunately this is only one step on the path to creating a seamless pool of shared physical resources that can be allocated dynamically to any application. There are three major barriers standing in the way of achieving this virtualization nirvana:

• Restricted application mobility across servers: Each application remains intimately tied to a specific physical server for storage, IO, and OS / binary compatibility. Moving an application across servers requires moving the storage and IO identity along with it, a highly complex and sensitive task. Without mobility across servers, resources appear as small, isolated islands of capacity, instead of one large resource pool.

• No consolidation of larger applications: A single physical server is often too small to consolidate several large applications, especially when the deployment is based on smaller scale out servers.

• Lack of mature management tools: The number of OS-level virtual machines is exploding, and along with it the cost and complexity of virtual machine administration. This problem is exacerbated by the lack of mature software tools offering infrastructure-wide monitoring and control of OS-level virtual machines

Delivering a Virtual Infrastructure

Network attached processing is a new approach to delivering a virtual infrastructure for middleware based applications. Network attached processing sets out to leverage existing application-level virtualization, such as the Java virtual machine (JVM) or .NET CLR, to cleanly separate the server and its OS-dependent functions (IO, storage, OS calls, and binary-dependent native code) from the virtual machine-dependent code (typically Java code). This approach completely liberates the virtual machine from any dependency on a specific piece of hardware – no storage dependency, no IO dependency, no OS dependency, and no binary dependency. In other words, the application virtual machine is completely virtualized from the underlying platform it is running on!

The virtual machine can now be moved transparently over the network to execute on an Azul compute pool, comprised of multiple high-end Azul compute appliances. The compute pool constitutes a virtual infrastructure, a seamless pool of resources that is shared among multiple applications, assigned at run-time based on SLA-determined policies, and can be re-allocated dynamically to another virtual machine every 10 milliseconds.

Each appliance offers enough fully coherent compute capacity – up to 384 processor cores and 256 GB of memory – to overwhelm the capacity requirements of any individual virtual machine, enabling efficient sharing of capacity within an appliance. Any application can be launched on any compute appliance, enabling optimization of resources across appliances. Best of all, resource management is automated by the Azul Compute Pool Manager, which provides centralized monitoring and policy-based resource controls. The end result: each virtual machine is guaranteed to receive the exact amount of resources it needs, the moment it needs it, with no manual intervention from system administrators.

For more information on this study as well as how network attached processing can help your organization with similar transaction-intensive application deployment challenges, please contact sales@azulsystems.com or call 650.230.6650.