1 / 9

Distributed Visualization

Distributed Visualization. Our use of this term extends its traditional meaning Distributed : Still aim to support geographically distributed users and collaborative communities Decentralized infrastructure : The infrastructure does not need to be centralized as in “compute” centers

mandy
Download Presentation

Distributed Visualization

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. Distributed Visualization • Our use of this term extends its traditional meaning • Distributed: Still aim to support geographically distributed users and collaborative communities • Decentralized infrastructure: The infrastructure does not need to be centralized as in “compute” centers • Shared infrastructure: The comp/storage nodes can be independent Internet computers

  2. Distributed Visualization

  3. Applications LoRS:runtime system L-Bone exNode IBP Physical Layer Data Replication: exNodes D4 D2 D3 D1 F1 F2 F3

  4. Network Functional Unit NFU is novel due to: weakened semantics and control of security-sensitive operations. PE RD RD/WR RD MEM Memory mapping Input Allocation Input Allocation Output Allocation IBP

  5. Scheduling • Depots: {P1,P2,…,Pm} Pi described by bw bi & computational power ci • Partitioned dataset {d1,d2,…, dn}, k-way replication • Vis only need one copy of each dj • (Optional) DM tasks Mij replicates dj on Pi Key Challenge: Resource performance varies over time !!!

  6. Dynamic Data Movement • Some data partitions are just “unlucky” to be on slow or heavily loaded servers • After fast depots are done with local tasks, can dynamically “steal” some slow “partitions”

  7. Results: the depots • Most depots used running on Planet-Lab • Workloads varied across servers and time • Realistic test of feasibility on shared, decentralized infrastructure

  8. Results: the data • Test data: 30 timestep of Tera-scale Supernova Initiative, 75GB in total • Provided by Tony Mezzacappa (ORNL) and John Blondin (ORNL) under the auspices of DOE SciDAC TSI project

  9. Results: the performance • 800x800 image resolution, 0.5 step size in ray-casting, per-fragment classification and Phong shading • With 100 depots, the average rendering time: 237 sec

More Related