Dolby Loboratories is developing encoding IP for high definition video encoding. During development of encoding algorithms engineers constantly compare the output of their latest algorithms, that process fixed sets of input data, with reference data. The workload for these simulations is highly unsteady and in-house resources are limited. Dolby decided to move the simulation workload to Penguin’s public on-demand cloud POD to take advantage of the ability to scale encoding jobs. As a result the runtime of a single job was reduced from 2 days to 2 hours, allowing Dolby engineers to explore more variations to their algorithms.
The use of POD is contributing significantly to making our development process more efficient
Gopi Lakshminarayanan, Director of Software Development, Dolby Labs
Organizations involved in IP and Software Development depend on HPC resources to increase product quality and productivity. Developers of encoding algorithms require quick turnaround of sample runs to verify accuracy. Developers of complex software packages take advantage of distributed compilations across multiple systems to build binaries. Organizations that provide digital content run compute intensive image processing jobs. The common issues development organizations face are :
- Project driven workloads that are typically uneven resulting in under- or over-provisioned resources
- Lack of dedicated administrators
- Selection of the right components to ensure balanced performance for applications
HPC Solutions for IP and Software Development
Penguin Computing has successfully delivered HPC solutions for software and IP development to many organizations such as Earthmine, Dolby Labs and Fluid Inc. We understand how to design systems for development environments, whether it is a workgroup cluster or a high performance Linux desktop for power users.
With Scyld ClusterWare Penguin also offers a complete, in-house developed HPC software stack that was designed for ease of use. Scyld ClusterWare has been fully qualified and tested with our hardware. Through our ISV partnerships and our own software expertise we can help with building and pre-installing open source and proprietary applications. Rather than just selling a cluster we deliver complete application optimized turnkey solutions that ‘just work’ and can be quickly moved to production.
If you are running short on computational resources Penguin on Demand (POD) is the solution for you. POD is a public HPC cloud that is accessible on a pay-as-you-go basis. It enables you to leverage the benefits of HPC without having to invest in an HPC infrastructure.