Title: Global Scale TeleImmersion Network Performance Activities Jason Leigh, Oliver Yu, Linda Winkler, Ala
1 Global Scale Tele-Immersion Network Performance
ActivitiesJason Leigh, Oliver Yu, Linda
Winkler, Alan Verlo, Tom DeFanti Yong-joo Cho,
Ray Fang, Javier Girado, Liujia Hu, Tomoko Imai,
Naveen Krishnaprasad, Michael Lewis, Ya Ju Lin,
Dave Pape, Kyoung Park, Chris Scharver, Brenda
Silva, Liang WangJosh Eliason, Jinghua Ge, Eric
He, Atul Nayak, Shalini Venkatamaran
2Common Characteristics of Teleimmersive
Applications
3Characterization of Tele-Immersive Streams
4Network Research
5Maximizing Bandwidth Utilization over Long Fat
Networks
- Even if QoS via DiffServ or IntServ is available,
it still does not solve the Long Fat Network
problem - Problem is small TCP window sizes (well known
problem but still no widely accepted solution) - On SGIs change in window size requires kernel
rebuild - Size of window should be set to current available
BW of the network - CAVERNsofts Parallel Socket Striping works well
but is considered irresponsible use of networks
664K Window SizeAmsterdam to Chicago
Bursty as max bw reached but performance is still
good
764K Window SizeCERN to EVL
Bursty as max bw reached but performance still
good
8Window Size EVL 1.85M, SARA 64K
EVL to SARA
When window size is large enough no real benefit
to using parallel sockets
Sending client determines the window size
SARA to EVL
9Window size EVL 1.85M, CERN 640K
EVL to CERN
Similar story at CERN
CERN to EVL
10Anomalies
- Theoretical BW from EVL to SARA is 100Mbps
- Netperf UDP shows reasonable performance
- EVL to SARA 85Mbps
- SARA to EVL 65Mbps (5 more hops via Abilene)
- Netperf and Parallel sockets TCP shows only
- 30Mbps
- Perhaps due to asymmetric tcp window size
settings? - Argument for UDP-based schemes?E.g. Forward
Error Correction
11Forward Error Correction scheme for low-latency
delivery of error sensitive data
- Transmit error correction data over high
bandwidth networks that can be used for
correcting UDP streams to achieve lower latency
than TCP but higher reliability. - Transmit error correction data to improve quality
of streamed video by correcting for lost packets. - Not intended for bulk data transfer but in light
of TCP results this might hold some promise.
12FEC Experiments
- EVL to SARA- Amsterdam (45Mb/s 100ms RT latency)
- Broader Ques
- Can FEC provide a benefit? How much?
- Tradeoff between redundancy and benefit?
- Specific Ques
- TCP vs UDP vs FEC/UDP
- How much jitter does FEC introduce?
- High thru put UDP vs FEC/UDP to observe loss
recovery
13FEC greatest benefit is in small packets. Larger
packets impose greater overhead. As redundancy
decreases FEC approaches UDP.
14(No Transcript)
15Packet Loss over UDP vs FEC/UDP between Chicago
Amsterdam
UDP
UDP
FEC
16Human Factors in Tele-Immersion
17Collaborative Coordination Experiments between
Chicago and Singapore
- CAVE to CAVE (STAR TAP)
- Audio via Phone call
- Scramnet (adjustable latency, 0 jitter)
- LAN Ethernet ( 10ms)
- Local ISDN ( 200ms)
- STAR TAP ( 250ms)
- Predict STAR TAP similar to performance over
ISDN
18Collaborative Coordination Experiments between
Chicago and Singapore
- 200ms RTT is the threshold where performance
begins to suffer - Roughly RTT to Asia. Results to Singapore similar
to local ISDN
200ms RTT with 0 jitter is same as 10ms RTT with
7ms jitter
19DiffServ Experiment 1
background
DiffServ
Bandwidth recovery good
EVL
100Mbps
100Mbps
x
25Mbps
80Mbps
Latency recovery good
x
42Mbps
42Mbps
ANL
x
x
100Mbps
100Mbps
Small packet loss
20DiffServ Experiment 2
background
DiffServ
Bandwidth recovery good
EVL
100Mbps
100Mbps
x
25Mbps
80Mbps
x
Latency recovery not good
42Mbps
42Mbps
ANL
x
x
100Mbps
100Mbps
Packet loss double
21Application of Research Results
- CAVERNsoft G2
- applications at iGrid 2000 in Yokohama
22Tele-Immersion MiddlewareThe CAVERNsoft G2
Toolkit
- G2 is C toolkit for building Tele-Immersive
applications with special emphasis on networking - Networking
- UDP, TCP, Multicast, HTTP.
- UDP reflector and multicast bridge.
- TCP reflector.
- Remote procedure calls.
- 32 and 64bit Remote file I/O.
- Parallel 32 64 bit TCP socket striping for high
throughput data delivery. - FEC library.
- Client/Server distributed shared memory
persistent database. - Threading, Mutual Exclusion.
- Built-in Instrumentation of networking services.
- QoS via GARA and MCSP underway.
23Tele-Immersion MiddlewareThe CAVERNsoft G2
Toolkit
- Audio streaming.
- Articulated Avatars.
- VR navigation.
- VR menus.
- Speech recognition with IBM ViaVoice.
- Collaborative application shell to jumpstart
development.
24TIDE
- Teleimmersive Data Explorer (TIDE)
- In collaboration with National Center for Data
Mining - General framework for collaborative visualization
of massive data-sets - Current data-set is ozone data from NOAA
25CIBRView
- Collaborative Image Based Rendering
Viewer(CIBRview) - In collaboration with Wes Bethel and Steve Lau at
Lawrence Berkeley Lab - Accesses volume data 512x256x256x 256 frames
40Gig data-sets - Generates image slices that are distributed to
collaborating clients. - Sent about 500 slices/files from Chicago to Japan
26Virtual HarlemUniversity of Missouri
Virtual Harlem Reconstruction of Harlem during
the Harlem Renaissance 1920-40
27Earthquake HypocentersSpace Physics Aeronomy
Research Collaboratory(U of Michigan)A
demonstration at Telecom 2000 and SC 2000between
Israel, Dallas, Chicago, Michigan
28Network Visualization
29QoS Internet Monitoring ToolQoSIMoto
30STAR TAP Network Visualization
31Future Work
- DiffServ and RSVP from EVL to CERN in
collaboration with NWU - Reliable UDP for high throughput bulk data
transmission - Integrated Collaboratory for Analysing Networks
(iCAN) iCAN-Monitor, iCAN-Visualize,
iCAN-Manage, iCAN-Active Test, iCAN-Collaborate