Compute Pools Feature Story

Compute Pools – Deep and Wide

The key to network attached processing is a massively scalable “compute pool”—a shared network resource accessible from any operating system or application server middleware. The compute pool, created by Azul Compute Appliances, can provide virtually unlimited compute resources to multiple Java platform-based applications without application level modification or binary compatibility requirements. It can be transparently deployed into an existing infrastructure, independently of the existing hardware and operating systems.

Handle Unexpected Requirements and Scale Your Business

Each Azul Compute Appliance contains up to 384 processor cores and 256 gigabytes of coherent shared memory in a flat symmetric multiprocessing (SMP) system that has been specifically optimized to run application virtual machine environments such as Java™ and J2EE™. Each appliance comes in a space-saving 11U rack-mountable enclosure enabling a single industry standard rack to house a compute pool with 1,280 processor cores (over 1 kilo cores!) and over 768GB of memory.

The Bottom Line—Reduced Complexity, Increased Utilization, Lower Cost
Accessible from multiple hardware, operating system, or middleware platforms, network attached processing solves the problem of heterogeneity. It also eliminates the need for capacity planning at the individual application level. Multiple applications can share processing resources available in the compute pool so that workload spikes are handled seamlessly across a data center. When calculating capacity, the entire compute pool is used in the calculation assuring that each application has the resources it needs while unlocking, consolidating, and thus minimizing the excess capacity that is required for peak loads.. The usual 5-to-1 peak-to-average ratio plus 100 percent safety margin is replaced by a much lower peak-to-average ratio for the whole pool, delivering what Azul Systems’ early field trials indicate as a 300 percent improvement in overall resource utilization.

Reacting to increased loads in real-time is a rare capability in data centers. An Azul compute pool addresses the problem in an architecturally efficient manner, enabling the allocation and re-allocation of systems resources within 10 miliseconds! This is especially important since an application spike in a transaction processing environment typically lasts between 30 and 100ms. Azul compute pools can actually respond to the spike as it occurs.

By tapping into the massive power of a compute pool, a Java application running on a host server is able to handle upwards of 30 times the workload possible with a single CPU traditional server host. This result is based on Azul Systems’ early field trials.

A Moment in History
The creation of network attached processing, designed to solve specific problems in today’s data centers, may well be the turning point for Java technologies, propelling the trends toward hardware and application consolidation, and ushering in a new era in application development in which developers are free to create massively scalable applications.

For more information on how network attached processing is delivering better SLAs, increased utilization of existing infrastructure and decreasing costs by 50%, download the latest analyst report, or contact us at either 650.230.6650 or sales@azulsystems.com.