Presentation on theme: "Access Grid: Australian activities ANU Internet Futures group and GrangeNet Grid Services Coordinator."— Presentation transcript:
Access Grid: Australian activities ANU Internet Futures group and GrangeNet Grid Services Coordinator
Australian AG nodes Currently around nodes Qld: UQ, QUT, JCU-T, JCU-C, CQU, GU NSW: U.Syd, UTS ACT: ANU(3), AARNet/GrangeNet Vic: VPAC, Monash, U.Ballarat SA: UniSA Expecting around 30+ by mid 2004.
AU-AGN’s Mix of technology selections Lots of all-Linux One all FreeBSD “standard” Windows/Linux mix All Windows Some PIGs, etc.
Developments in AG-AU Several groups working on shared visualisation applications Some groups working on instrument integration Electron microscopy Telescopes At least one group working on usability, performance and functionality Mine
A separate talk: The Good, The Bad, and The Ugly HCI/GUI Issues, A/V localisation, Large group representation Audio/Video tracking of (moving) targets, A/V Synchronisation Whiteboards, tablets, laser pointers Centralised and automated node operation High quality A/V Performance Measurement, Analysis, Tuning Stressing Reliability/Availability Holography Haptics Non-flat and immersive projection Rear-wall projection Aids remote interaction Intelligent gateways H.323, SIP Transcoding Source selection Streaming Recording and playback Support non-AG spaces, non-GrangeNet bandwidth Layered codecs Stereo to 2-D Security Access, encryption, Network Services Smarter (reliable?) multicast, QoS wireless devices Additional interaction modes Linkage of Venues, Services and applications Linkage to other (any!) Grid Services computing, data, instruments, visualisation Visualisation Shared or independent views Data browser Sharing, discussion, annotation Computational Steering Instrument Consoles Screen Scraping vs application awareness It’s important! It’s independent of the space! Physical Security Use of public spaces for AGN, tempting equipment Use AG equipment (cameras, mikes) for automated, intelligent security out of hours. Use other sites as your security guard (follow the sun security)
~4 staff plus 4 students Building 3 AG nodes, another 3-4 on order/design Strong user focus – astronomers, medical school, biologists, administrators Strong interest in visualisation (3D) and computational/data interfaces. Attacking several problems now Performance (analysis/measurement) Connectivity (multicast/unicast) UI (managing video/audio) Integration (H.323, SIP) User friendliness, node operator removal Hardware integration (electronic whiteboards)
Performance measurement RTPreplicate Stress your display/audio machines One stream goes in, N come out Use feedback from rtcp to benchmark Derive a meaningful number Full throughput test network, interfaces, CPU/RAM/graphics bus
Connectivity Multicast isn’t everywhere Bandwidth isn’t everywhere Multicast/unicast bridges don’t scale well in management Host must reconfigure all the time RCbridge and rcb-forward client
RCBridge http is scriptable – command line clients Can build a P2P network of unicast fowarders, e.g. for failover
User Interface Vic (video) and rat (audio) old, unfriendly Few new codecs (DV, MPEG-x, MJPEG, HDTV, …) And don’t scale all that well No lip-synch no mbus No audio-localisation No “remote-site UI management” Manageability is a problem Major effort of node operator No ability to scale codecs across clusters Performance and reliability advantages
VP – a new start Plug-able architecture Codecs Filters (pre/post) E.g. captions Devices Use of OpenGL capabilities Smooth (not just quantised) scaling, done in hardware Much more flexible user interfaces possible Audio, lip-synch, localisation, UI, site-grouping Starting with a decode app first
Integration with H.323 H.323 is wide-spread, AG is not (yet) Bridging between two worlds Analogue VRVS Silver23 Open source, software gateway Some MCU-like functionality Hardware constrained, but clusterable http managed, very flexible stream selection Provides an easy path into our VoIP and PSTN Integrate with telephony with no additional hardware Your mileage may vary…
Hardware, and Humanware Integrate electronic pen-tracking whiteboard Astronomers want blackboard/chalk integration! E-beam and Mimio are excellent Node operator is too expensive Design nodes to be remotely managed Hardware integration with out-of-band interfaces Maintain rapid user support in rooms Generalised application sharing is a BIG issue. Treat VGA as a new kind of video source, separate channel for mouse/keyboard events