1 / 3

1. Project Goal (SISA Omni-OS)

1. Project Goal (SISA Omni-OS). Project Goal - Develop an Operating System that will scale to 80% efficiency at 1000+ cores by 2015 and support a transition path from Linux and GNU based ecosystem (2011) 50%@48 --> (2013) 80%@100 cores incrementing up to 80%@1000 cores by 2015

Download Presentation

1. Project Goal (SISA Omni-OS)

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. 1. Project Goal (SISA Omni-OS) • Project Goal • - Develop an Operating System that will scale to 80% efficiency at 1000+ cores by 2015 and support a transition path from Linux and GNU based ecosystem • (2011) 50%@48 --> (2013) 80%@100 cores incrementing up to 80%@1000 cores by 2015 • - Necessary system S/W to realize the benefits of manycore processors and enable future “killer applications” that require massively concurrent execution • - 1000-core processors expected to be commercially viable by 2015. Measurements on existing state-of-the-art Operating Systems (e.g., Linux, MIT fos) indicate < 10% efficiency at 1000 cores Benefit • - Realize processing and power saving potential of future homogeneous and heterogeneous processors • - Enable capabilities not viable with todays Mobile Operating System solutions (real-time, QoS, flexibility) • - Provide a technology path to transparent “leap frog” cloud computing solutions • - Position Samsung to have its own differentiating OS solution (non-reliance on Google and Microsoft) • - Provide a OS solution for SAIT in-house advanced processor technology (refer to SAIT System Arch Lab) Need - OS and System S/W solution for manycore processor technologies - Enable platform (H/W + S/W) differentiation through application performance (scaling and real-time), portability, reliability and security (refer to SAIT/ICL LEGO-OS project) - Fundamental technology with broad business impact across existing CE and future business areas - [Ultimately] Open Source driven to capture existing GNU/Linux communities and ecosystems

  2. 2. Project Approach • Project Approach • Key success factors: • continuation of Moore’s law circuit fabrication scaling (e.g., planar < 10nm, 3D) and commercial realization of 1000-core processor • advent of alternatives to CMOS may drive return to frequency scaling as principal differentiator • Apple/Microsoft/GNU and Linux community are slow to move to scalable architecture for manycore (Samsung can develop both S/W and H/W technology at the same time [integration with SAIT-SAL Adv. Processor work]) • Unique approach • distributed (on- and off-chip) micro-kernel architecture that eliminates global synchronization (scalability limiter) • holistic view that integrates parallel programming (compiler/runtime) with OS capabilities Competition - GNU Hurd microkernel (http://www.gnu.org/software/hurd/) will likely replace monolithic kernel in the Linux community; this architecture lends itself to the development of a scalable solution - other academic manycore Operating Systems (MIT -FOS, Microsoft/ETH Barrelfish (project ended), UC Berkeley Tessellation); these OSesare early prototypes and data indicates poor scalability at 1000 cores - unknown competition by Apple/Google/Microsoft Approach • - rework existing L4 microkernel (TU Dresden) to eliminate scaling limiters and integrate manycore-specific features (e.g., partitioned scheduling) • - build a new user-land “personality” based on ideas from Genode Labs that provides fine-grained resource management; current Genode Labs solution don’t scale • - integrate new, innovative OS designs that focus on scalability on manycore (e.g., distributed physical memory management • - collaborations with Purdue University/Genode Labs

  3. 3. Technology (List needed technologies)

More Related