Super Duper Networking Transforming Supercomputing from the point

A Typical Data Correlation and Visualization Pipeline Data Source Correlate/Filter Render Display Data Source

Experiment to Use Inexpensive Photonic Switches as an alternative to traditional million $ routers

In Collaborative Work, Data or Visualization needs to be Distributed to Collaborating Sites Data

Photonic Multicast Service Glimmerglass Reflexion Photonic Multicast-capable Switch Electronic Visualization Laboratory (EVL) University of

Photonically Multicasting a Visualization Challenges: • Need to augment traditional Routing and Wavelength Assignment

1 st Step: Realize Local Area Photonic Multicasting (Visit Booth R 2935 to learn

Quiz: Guess the Mystery Computer with the enormous bandwidth but tiny caches… 2. 4

Slides: 8

Download presentation

Super. Duper. Networking Transforming Supercomputing …from the point of view of Large Scale Visualization and Collaborative Work Jason Leigh spiff@evl. uic. edu Electronic Visualization Laboratory (EVL) University of Illinois at Chicago

A Typical Data Correlation and Visualization Pipeline Data Source Correlate/Filter Render Display Data Source Render + Display Data Source Correlate Render + Display Data Source + Correlate Render Display Things to notice: - Pipelines are static for long periods of time. NOT like web surfing- so…. - Routing is not crucial. - Program code is tiny compared to volume of data processed. Caching won’t help much- so…. - Need to stream lots of data through fast concurrent pipelines! - Need pipelines to be optimized from end to end. Electronic Visualization Laboratory (EVL) University of Illinois at Chicago

Experiment to Use Inexpensive Photonic Switches as an alternative to traditional million $ routers to provide application-controlled deterministic network paths/pipelines. Long haul link The cross connections are applicationprogrammable. Protocol & data rate independent Calient / Glimmerglass at Star. Light & EVL Electronic Visualization Laboratory (EVL) University of Illinois at Chicago

In Collaborative Work, Data or Visualization needs to be Distributed to Collaborating Sites Data Source Correlate Render Display Data Source Render + Display Data Source Render Display Data Source Correlate Render + Display Data Source + Correlate Render Display Electronic Visualization Laboratory (EVL) University of Illinois at Chicago

Photonic Multicast Service Glimmerglass Reflexion Photonic Multicast-capable Switch Electronic Visualization Laboratory (EVL) University of Illinois at Chicago

Photonically Multicasting a Visualization Challenges: • Need to augment traditional Routing and Wavelength Assignment algorithms to consider photonic multicast constraints. • Need extreme speed reliable multicast protocol Electronic Visualization Laboratory (EVL) University of Illinois at Chicago

1 st Step: Realize Local Area Photonic Multicasting (Visit Booth R 2935 to learn how this is done on the Opt. IPuter) Electronic Visualization Laboratory (EVL) University of Illinois at Chicago

Quiz: Guess the Mystery Computer with the enormous bandwidth but tiny caches… 2. 4 GB/s from main memory to graphics (Today’s AGP 8 X is at 2. 1 GB/s) 48 GB/s! (Today’s Quadro FX 3000 only has 27 GB/s) Tiny caches SIF IPU 128 bit bus DMA Memory 32 MB FPU 16 K cache 300 Mh. Z MIPS 3 Graphics Synthesizer 4 MB VU 0 VU 1 GIF Vector processors with several parallel pipelines For distributed, collaborative large scale data visualization, we need a version of this that extends to wide area environments. Electronic Visualization Laboratory (EVL) University of Illinois at Chicago