R Review of DVTSplus System i f DVTS l S t Congxiao Bao CERNET T i CERNET, Tsinghua University h U i it 2011-8-24
Participants Today Participants Today • 1.Main Venue in Delhi 1 M i V i D lhi • 2. Nation TaiWan University hospital ,Taipei ,(NTU) • 3 National Center for High • 3. National Center for High- performance Computing (NCHC), Taiwan (NCHC) • 4.Kyushu University Hospital, Fukuoka, Japan (KUH) 4.Kyushu University Hospital, Fukuoka, Japan (KUH) • 5.Masaryk Hospital, Usti, Czech (MNUL) • 6.CESNET HQ, Prague, Czech (CHQ ) , g , ( ) • 7.University of Malaya, Malaysia (UOM ) • 8.Asian Institute of Gastroenterology, Hyderabad, India (AIG) • 9. Tsinghua Univeristy Beijing ,China (THU)
Outline Outline • System description • Past experiences Past experiences • Statistics analysis from past usage • Summary and future work
What is DVTSplus What is DVTSplus • Extension of DVTS system for multi-sites communication – Keep DVTS video/audio quality – More flexible management dealing with More flexible management dealing with network sensitivity, time-consuming testing, automatic scheduling and limitation of number t ti h d li d li it ti f b of participants
Look in deep Look in deep • Many clients, distributed servers (work together as a M li t di t ib t d ( k t th whole MCU) • No decoding, just forwarding g, j g • Standard DVTS video/audio, combined with small video/audio, represented by different channels • Different application scenarios determine different Different application scenarios determine different signaling • Signaling system determine which channel to forwarding g g y g • Two basic modes include Chairman mode and discussion mode • The number of participants are limited by hardware • The number of participants are limited by hardware configurations (CPU/memory/bandwidth of Ethernet card etc.)
From user’s operation From user s operation • Same local hardware setup as WIDE DVTS • Web-based to access the system http://conf dvtsplus org http://conf.dvtsplus.org • Just “click” on the website to achieve your own control in discussion mode
To look back To look back • First release First release – 2010.6 • First trial (version 0.3) First trial (version 0.3) – 2010.8 (30th APAN, VN) – 9 sites in Asia area • Second trial (version 1.0) S d t i l ( i 1 0) – 2010.11 (2010 Fall Internet2 meeting) – 7 sites in Asia and North & South America • Third trail (version 1.3) – 2011.2 (31th APAN, HK) – 10 sites in Asia-pacific area 10 sites in Asia pacific area • Forth trial (TERENA 2011, Prague) – 14 Sites in Asia-pacific and Europe
30th APAN 30th APAN
2010 Fall Internet2 member meeting 2010 Fall Internet2 member meeting
31th APAN 31th APAN
TNC2011 TNC2011
Many thanks to Many thanks to • Great technical supports from engineers of network and video/audio in each station • Valuable comments, feedbacks and suggestions from doctors as well suggestions from doctors as well
what’s what s • What you saw in past demo • What are the reasons behind What are the reasons behind • What we have done since then
First trial First trial • In the demo – One site with blinking small video (no large g ( g one) and much noises, made the general feeling in the main venue was not good g g – One site had to adjust the sound volume from time to time to get acceptable quality time to time to get acceptable quality – Chairman in the venue had some trouble in finding who will be the next questioner during finding who will be the next questioner during the Q&A
First trial First trial • Reasons behind R b hi d – Local network problems, the whole system could not be able to do anything to improve – Not so skillful on system adjustment, the whole system didn’t provide echo cancellation functions, the quality of voice largely depends on the local engineer's skills th l l i ' kill – Many participants shown in the many small videos, difficult to be recognized at very short id diffi lt t b i d t h t time
Second trial Second trial • In the demo – Echo still can be found even the local engineer is skillful – The Voice quality from the standard DVTS The Voice quality from the standard DVTS can be affected by a very small packet loss
Second trial Second trial • Reasons behind – Still no professional echo-cancellation p function in the system – Standard DVTS have video&voice build-in Standard DVTS have video&voice build in together • New requirements N i t – HD – Distributed server
Third trial Third trial • In the demo – Almost perfect p • More new requirements – Recoding R di – Separate screens
Fourth trial Fourth trial • In the demo – Unstable video quality between Beijing and q y j g Prague when enable a server in Prague – Small videos can not be seen all the time so Small videos can not be seen all the time, so the operator is “blind” – one stations can not hear Beijing s voice one stations can not hear Beijing’s voice – Discontinuous voice between Beijing and P Prague? ?
Fourth trial Fourth trial • Reasons behind – Jitter and packet loss in Beijing-Europe link p j g p – The resolution in the Prague venue is very low (600*800) low (600 800) – Maybe Some problem in local PC (not completely locate) completely locate)
(7) (6) (5) (4) (3) (2) (1)
What we have done What we have done • Since 1st trial Si 1 t t i l – Online testing server running 24 hours in a separate session session – Create chairman mode besides discussion mode – Chairman can moderator the video/audio transmission t i i – Audio control to avoid echo • Only chairman ‘s voice and one selected voice are in the y channel – Queuing mechanism for questioner – Distributed server without load balancing Distributed server without load balancing – Configurable for firewall settings – Network conditions monitoring – Web-based scheduling system
Network condition monitoring Network condition monitoring
In the plan In the plan • Echo-cancellation • Load balancing for distributed server Load balancing for distributed server • HD • IPv6 support
Thank you y
Recommend
More recommend
