Click to edit Master title style Click to edit Master title style - PowerPoint PPT Presentation

Click to edit Master title style Click to edit Master title style Click to edit Master title style ? Click to edit Master text styles ? ? Click to edit Master text styles ? Click to edit Master text styles TAROC: TCP-Aware RObust ` Second level ` ` Second level ` Second level Header Compression Scheme ? ? Third level ? Third level ? Third level ` ` Fourth level ` Fourth level ` Fourth level ? Fifth level ? Fifth level ? Fifth level ? http://www.dmn.tzi.org/ietf/rohc/draft-ietf-rohc-tcp-taroc-01.txt HongBin Liao, Qian Zhang, Wenwu Zhu, Ya-Qin Zhang Microsoft Research 1 1 1 1 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

Outline Click to edit Master title style Click to edit Master title style Click to edit Master title style ? ? Click to edit Master text styles ? Click to edit Master text styles ? Click to edit Master text styles Basic concepts of TAROC robustness of TAROC (window-based LSB encoding) ` Second level ` Second level ` Second level ` efficiency of TAROC (TCP congestion window tracking) ? Third level ? Third level ? Third level ? ` Fourth level ` ` Fourth level ` Fourth level Improvements in this version ? Fifth level ? Fifth level ? ? Fifth level more accurate for TCP congestion window tracking more efficient for bulk data transmission adjust packet header format Tracking / compression performance analysis 2 2 2 2 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

Window-based LSB encoding Click to edit Master title style Click to edit Master title style Click to edit Master title style Seqno variation distribution Ackno variation distribution ? ? Click to edit Master text styles ? Click to edit Master text styles ? Click to edit Master text styles ` Second level ` Second level ` Second level ` ? ? Third level ? Third level ? Third level ` Fourth level ` ` Fourth level ` Fourth level ? ? Fifth level ? Fifth level ? Fifth level 0-100 100-200 200-300 300-400 400-500 500-600 600-700 700-800 800-900 900-1000 1000-1100 1100-1200 1200-1300 0-100 100-200 200-300 300-400 400-500 800-900 1300-1400 1400-1500 1500-1600 1600-1700 1700-1800 1800-1900 1900-2000 2000-2100 2100-2200 2200-2300 2300-2400 2400-2500 2500-2600 600-700 1400-1500 500-600 900-1000 1000-1100 1100-1200 2600-2700 2700-2800 2899-2900 2900-3000 3000-3100 3100-3200 3200-3300 3300-3400 3400-3500 3500-3600 3600-3700 3700-3800 3800-3900 1200-1300 1300-1400 700-800 others 3900-4000 4000-4100 4100-4200 4200-4300 4300-4400 4400-4500 4500-4600 4600-4700 4700-4800 4800-4900 4900-5000 others 03/11/2001 - 03/17/2001 SDSC : San Diego Supercomputer Center 3 3 3 3 NLANR network traffic packet header traces Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

Tracking-based TCP congestion Click to edit Master title style Click to edit Master title style Click to edit Master title style window estimation Reconstruct the congestion control behavior of TCP sender at the ? Click to edit Master text styles ? ? Click to edit Master text styles ? Click to edit Master text styles compressor ` Second level ` Second level ` Second level ` Adjust congestion window size in each state based on seqno /ackno ? Third level ? Third level ? Third level ? ` ` Fourth level ` Fourth level ` Fourth level ������������� ����������� ��������� ? Fifth level ? ? Fifth level ? Fifth level ������������� ����������� ���������� � � � ����� � � � � � � �������� � � 4 4 4 4 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

Several issues of TAROC version 0 Click to edit Master title style Click to edit Master title style Click to edit Master title style ? Click to edit Master text styles ? ? Click to edit Master text styles ? Click to edit Master text styles Accuracy of tracking-based TCP congestion window estimation ` Second level ` Second level ` Second level ` Ackno-only window tracking ? ? Third level ? Third level ? Third level Receiver advertised window driven case ` ` Fourth level ` Fourth level ` Fourth level Multiple link case ? Fifth level ? Fifth level ? Fifth level ? Efficiency of window-based LSB coding Bulk file transmission usually has fixed payload size Irregular seqno / ackno for other traffics More realistic and abundant simulation conditions Simulation topologies (multiple bottleneck, multiple link) Simulation BER (10 -5 ~ 10 -8 ) 5 5 5 5 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

Improvement in congestion window estimation Click to edit Master title style Click to edit Master title style Click to edit Master title style ? ? Click to edit Master text styles ? Click to edit Master text styles ? Click to edit Master text styles new packet comes in ` Second level ` Second level ` Second level ` Seqno / Ackno-based Congestion window increase Congestion window increase ? Third level ? Third level ? Third level ? congestion window estimation seqno/ackno-based congestion window seqno/ackno-based congestion window estimation ` ` Fourth level ` Fourth level ` Fourth level estimation Congestion window shrunk Congestion window shrunk ? Fifth level ? ? Fifth level ? Fifth level RTT-smoothed rate seqno/ackno-based congestion window seqno/ackno-based congestion window estimation estimation estimation RTT-smoothed rate estimation RTT-smoothed rate estimation (TCP linear / sub-linear increasing) (TCP linear / sub-linear increasing) Estimated congestion window adjustment new packet comes in 6 6 6 6 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

RTT-smoothed rate estimation Click to edit Master title style Click to edit Master title style Click to edit Master title style ? Click to edit Master text styles ? Click to edit Master text styles ? ? Click to edit Master text styles RTT estimation ���� Packet clustering distribution ` Second level ` ` Second level ` Second level > µ × ���� ��� ? Third level ? Third level ? ? Third level ` sub-clustering ` near-uniform distribution ` Fourth level ` Fourth level ` ` Fourth level Inter-RTT variation judgment ? Fifth level ? ? Fifth level ? Fifth level RTT-smoothed rate calculation ∑ = ����� i i ���� ��� ��� Congestion window shrunk ���� ��� ��� ��� ��� Check the conflict between rate variation and cwnd variation 7 7 7 7 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

Compressed TCP/IP header structure Click to edit Master title style Click to edit Master title style Click to edit Master title style Almost the same as the one ���� ? Click to edit Master text styles ? Click to edit Master text styles ? ? Click to edit Master text styles 0 in version 1 except for ����������������� 1 ��������� F ����������������������������� ` Second level ` ` Second level ` Second level Bulk data transmission (fixed 2 payload size) <F+WAS fields> ������������ ? ? Third level ? Third level ? Third level 3 W-LSB encoding applied to ` Fourth level ` Fourth level ` ` Fourth level ��������� ��������������������� Window, Acknowledgement �������� ������� ? ? Fifth level ? Fifth level ? Fifth level Number, Sequence Number and �������������� �������� IP-ID ������� TS value, TS echo reply in TCP ��� ��������� Timestamp option ����� ������� �������� Delta coding applied to �������� SACK option ��������� �������� �������������������� 8 8 8 8 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt

WAS field Click to edit Master title style Click to edit Master title style Click to edit Master title style ? Click to edit Master text styles ? Click to edit Master text styles ? Click to edit Master text styles ? PT C_WIN C_ACK C_SEQ C_IP_ID PADDING ` Second level ` Second level ` Second level ` F PT C_WIN C_ACK C_SEQ C_IP_ID length of value length (bits) length (bits) length (bits) length (bits) WASI (bytes) ? Third level ? Third level ? ? Third level 1 xxx 3 4 4 3 1 ~2 ` ` Fourth level ` Fourth level ` Fourth level 0 010 5 6 6 5 2~4 ? Fifth level ? Fifth level ? Fifth level ? 0 011 8 9 9 8 3~5 0 100 13 14 14 13 4~8 0 00x 10 11 11 3 2 ~5 0 101 13 14 14 6 3~7 0 110 17 18 18 10 4~9 0 111 21 22 22 14 5~11 9 9 9 9 Mar 2001, 50 th IETF draft-ietf-rohc-tcp-taroc-01.txt