1 / 11

The Gathering Storm: OSs and VMMs meet Many-Core

The Gathering Storm: OSs and VMMs meet Many-Core. Mike Swift University of Wisconsin-Madison. What will change?. Lots of execution contexts Some extra, but not many High application demand for parallelism Low-overhead communication and synchronization Heterogeneous processors

hedia
Télécharger la présentation

The Gathering Storm: OSs and VMMs meet Many-Core

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. The Gathering Storm:OSs and VMMs meet Many-Core Mike Swift University of Wisconsin-Madison

  2. What will change? • Lots of execution contexts • Some extra, but not many • High application demand for parallelism • Low-overhead communication and synchronization • Heterogeneous processors • Statically and dynamically • Closely-coupled systems • Shared caches, buses

  3. Challenges • Single thread performance has flattened • OS overhead matters more • Must reduce single-thread overhead • Must perform more work asynchronously

  4. Opportunities • Hardware support for synchronization and fast communication • Fine-grained parallelism within the kernel • Transactional memory to simplify kernel coding

  5. Opportunities • Extra hardware threads for background tasks • Monitoring • Optimization • Redundant execution

  6. Challenges • Scheduling matters again • We have to read all those gang scheduling papers • OS must to manage diverse resources • Set of resources may change • Threads may interfere • Application demand may change

  7. Challenges • Dynamic hardware • OS or VMMs must tolerate CPU addition, removal, change in properties

  8. Challenges • Scalability • Existing SMP, ccNUMA techniques difficult to apply, reduce single-thread performance • Partitioning approaches limit resources to applications

  9. Opportunities • Possible new OS structures for scalability • Adopt supercomputer approach with OS nodes, compute nodes • Multiple quasi-independent kernels • Independent kernels with DSM for large applications

  10. Are we up to it?

  11. Challenges • Interference • OS must account for interference between threads • Hardware must provide measurements of interference

More Related