ºÝºÝߣ

ºÝºÝߣShare a Scribd company logo

Why geni

?Download as PPTX, PDF?
?
David Lary
David Lary

This document discusses distributed collaborative scientific visualization using GENI, Ignite, and big scientific data. It finds that for collaboration around big data, all interactions are limited by size and speed, with data size and server proximity determining performance. The optimal setup places replicated database servers in close proximity to users to enable collaborative visualization of big data from any device anywhere with response times under 150 milliseconds.

1 of 18
Distributed Collaborative Scientific Visualization GENI, Ignite, and Big Scientific Data
Visualize Big Data Fast Any Device Anywhere Collaboratively
Size and Speed Target: 0.15 seconds * * N. Tolia, D. G. Andersen, and M. Satyanarayanan. Quantifying interactive user experience on thin clients. Computer, 39(3):46¨C52, 2006.
Why geni
Why geni
Conclusion Same Building is Best Same ¡°City¡± (State) is OK Same Continent Won¡¯t Fly One Big Server = World Wide Wait
Meaning For Collaboration Server must be in same location as user Distributed Collaboration: users can be anywhere Means server requirement: servers must be everywhere
System Setup < 10 ms, > 100 mb/s Replicated Databases
Global Picture
Distributed Collaboration Around Big Data ¡°A World Where Distance is Eliminated¡± Experts Around the World Interacting With Data Visualizations as Easily as if They Were in the Same Room Previously: Expensive Hardware (OptIPortal, CAVE) over Expensive, Special-Purpose Networks GENI: Any Device, Anywhere, Anytime, through a Web browser
Collaboration Around Big Data All About Size and Speed - Data Set: 4 million points per month - 100 MB/month Too Much Data For Laptop Way Too Much for Tablet/Phone/Netbook Need Server Close to User How Close Depends on Bandwidth
Size and Speed Task: Draw 30,000 circles in 160 milliseconds World at 100-km resolution Quarter-continent at 10-km resolution Requirement: 160-milliseconds User studies show dropoff beyond that. Question. Can we do that from: Server on campus, Server in city, Server on continent, Single Server for World?
Size And Speed Server in Building Server in City Server on Continent Worldwide Server Request Time 1 5 50 250 Fetch Time 20 20 20 20 Transmit Time 8 30 300 1500 Draw Time 100 100 100 100 Total 129 155 470 1870 Time To Draw 30,000 Points in milliseconds. Goal: 150 ms
Deployment Options Sparse Interaction/Small data Rich Interaction/Big Data Powerful Client Anything Desktop Application Any Client Classic Cloud GENI/Ignite Cloud
Size and Speed (30,000 points) 0 150 50 35 41 141 Using GENI 0 50 80 140 240
Size and Speed (10,000 points) 0 150 50 35 41 141 Using GENI 0 50 80 140 240
Size and Speed (30,000 points) 0 150 50 35 41 141 Using GENI 0 50 80 625 525 Using Standard Internet
Size and Speed (10,000 points) 0 150 50 15 20 50 Using GENI 0 50 60 120 170

More Related Content

Why geni

  • 2. Visualize Big Data Fast Any Device Anywhere Collaboratively
  • 3. Size and Speed Target: 0.15 seconds * * N. Tolia, D. G. Andersen, and M. Satyanarayanan. Quantifying interactive user experience on thin clients. Computer, 39(3):46¨C52, 2006.
  • 6. Conclusion Same Building is Best Same ¡°City¡± (State) is OK Same Continent Won¡¯t Fly One Big Server = World Wide Wait
  • 7. Meaning For Collaboration Server must be in same location as user Distributed Collaboration: users can be anywhere Means server requirement: servers must be everywhere
  • 8. System Setup < 10 ms, > 100 mb/s Replicated Databases
  • 10. Distributed Collaboration Around Big Data ¡°A World Where Distance is Eliminated¡± Experts Around the World Interacting With Data Visualizations as Easily as if They Were in the Same Room Previously: Expensive Hardware (OptIPortal, CAVE) over Expensive, Special-Purpose Networks GENI: Any Device, Anywhere, Anytime, through a Web browser
  • 11. Collaboration Around Big Data All About Size and Speed - Data Set: 4 million points per month - 100 MB/month Too Much Data For Laptop Way Too Much for Tablet/Phone/Netbook Need Server Close to User How Close Depends on Bandwidth
  • 12. Size and Speed Task: Draw 30,000 circles in 160 milliseconds World at 100-km resolution Quarter-continent at 10-km resolution Requirement: 160-milliseconds User studies show dropoff beyond that. Question. Can we do that from: Server on campus, Server in city, Server on continent, Single Server for World?
  • 13. Size And Speed Server in Building Server in City Server on Continent Worldwide Server Request Time 1 5 50 250 Fetch Time 20 20 20 20 Transmit Time 8 30 300 1500 Draw Time 100 100 100 100 Total 129 155 470 1870 Time To Draw 30,000 Points in milliseconds. Goal: 150 ms
  • 14. Deployment Options Sparse Interaction/Small data Rich Interaction/Big Data Powerful Client Anything Desktop Application Any Client Classic Cloud GENI/Ignite Cloud
  • 15. Size and Speed (30,000 points) 0 150 50 35 41 141 Using GENI 0 50 80 140 240
  • 16. Size and Speed (10,000 points) 0 150 50 35 41 141 Using GENI 0 50 80 140 240
  • 17. Size and Speed (30,000 points) 0 150 50 35 41 141 Using GENI 0 50 80 625 525 Using Standard Internet
  • 18. Size and Speed (10,000 points) 0 150 50 15 20 50 Using GENI 0 50 60 120 170