Victoria Partnership for Academic Computing chooses Linux

An enterprise whose business processes-integrated end-to-end across the company and with key partners, suppliers and customers-can respond with speed to any customer demand, market opportunity or external threat.

In 2000, a consortium of seven universities in the Australian state of Victoria established the Victorian Partnership for Advanced Computing (VPAC), a nonprofit agency that provides high-performance computing (HPC) services in the fields of computational engineering, computational software development, geospatial sciences, grid computing and the life sciences. Clients include local research hospitals, government agencies and automotive giant General Motors.

VPAC computing projects range from aggregating colorectal cancer databases and checking large datasets for marine research to analyzing oil slosh and profiling consumer buying habits. Additionally, VPAC recently received government funding from the Australian Partnership for Advanced Computing (APAC) to help establish a national computational grid.

Anticipated workload growth

In 2004, VPAC aggregated high-performance computers around Australia, adopting international middleware standards and deploying grid portals to support applications of local interest.

Although the overall grid project is in its infancy, the company expects demand for its applications to increase significantly over the next year. Because the risk of service interruption increases with more servers connecting to the grid, high availability is a critical priority for VPAC.

'With a large computer network running huge jobs, downtime can be catastrophic,' says David Bannon, VPAC systems manager. 'We run some very long jobs-if a server goes down, it can mean months of wasted compute time even if that server is only down for a few minutes.'

Initially, the demands of grid processing heavily taxed VPAC's infrastructure performance. The company's aging Compaq AlphaSC supercomputer was expensive to maintain and operate, and it was performing well below current industry standards. VPAC management decided it would be less expensive to purchase a new supercomputer with optimal price/performance than to continue operating their current AlphaSC machine.

IBM server performance

After considering solutions from a number of competitors, including Sun Microsystems and Hewlett-Packard, VPAC replaced its AlphaSC supercomputer with a high-performance IBM Linux cluster solution.

Because of the organization's commitment to open standards, most of the systems in the VPAC infrastructure already ran the Linux operating system, leading the company's IT team to seek a new system that would also run Linux. 'We wanted the flexibility of open standards-based computing, and we had an existing Linux skill set that we wanted to maximize,' says Bannon.

VPAC purchased 37 IBM OpenPower 720 servers, which feature the SUSE LINUX Enterprise Server (SLES) V9 for PowerPC operating system. Thirty-six of the OpenPower 720 servers are clustered into computing nodes to support VPAC's HPC workload, and one server is partitioned separately for head and management nodes. The new IBM Linux cluster augments the client's two existing IBM Cluster 1350 servers, giving VPAC three IBM Linux cluster environments.

'We liked how the OpenPower system is tuned for Linux,' says Bannon. 'The system easily supports our existing applications, which gives us high performance at a price point that can help us reduce our acquisition costs, lower our total cost of ownership and simplify our IT infrastructure.'

On Demand Business Benefits

- Significant gain in computing power over the Compaq AlphaSC

- Capacity to pursue new clients and to participate in building a national computing grid

- Ability to handle large workloads running at 80 to 90 percent capacity with minimal downtime

- Lower costs through reduced power consumption, which VPAC management believes may help achieve return on investment (ROI) within four years

Linux-tuned OpenPower systems

Built on the IBM Power Architecture instruction set featuring flexible, stable IBM POWER5 technology, the OpenPower system is designed and engineered for reliability, availability and serviceability (RAS) from the ground up.

Advanced OpenPower RAS capabilities such as First Failure Data Capture and Dynamic Processor Deallocation help keep systems running, while the Linux kernel leverages IBM POWER performance features to help improve uptime and scalability. Leveraging OpenPower system features, VPAC has achieved 80 to 90 percent processor utilization.

Also, according to Bannon, the system has had minimal downtime in the eight months since VPAC starting running it. 'This is great performance-the kind of performance we require to support our customers and to establish a reliable portal in a grid computing environment,' he says.

With the OpenPower 720 systems at the center of its IT infrastructure, VPAC has improved its client responsiveness and boosted availability for clients' demanding workloads, providing users with a significant increase in computing power over its previous Compaq AlphaSC solution.

Says Professor Bill Appelbe, founder of VPAC, 'As a result of these performance gains, we've been able to improve overall client service levels. Jobs run much more quickly, and we have much shorter user queues. Our new system is giving us outstanding scalability as well, allowing us to add more concurrent users without affecting performance.'

Low energy costs

In addition to generating impressive performance figures, the OpenPower 720 Linux cluster has been extremely dependable and energy efficient compared to the Compaq AlphaSC.

'This OpenPower system is yielding substantial savings in energy costs. Because of the Power5 architecture's lower power consumption, we enjoy lower peripheral costs, such as the cost of air conditioning,' says Bannon. 'Over a few years' time, these can add up to substantial savings.'

Applebee is particularly impressed with the IBM's support and service, which enables his organization to focus on meeting the needs of its growing client base. 'Obviously, IT is intrinsic to our business and the services we offer clients. However, with IBM's support, we can feel confident about delivering the level of service users require.

For example, often IBM detects hardware issues before our clients experience a problem, so we can continue running work on the specific node until the job is complete, and then shut down the node and reroute other jobs.

The transition between nodes is completely transparent to users, so it helps us minimize downtime. We always knew we could count on IBM technology and support, and our choice has been validated by our growing number of satisfied clients.'