Improving Slow Start In Kwan Yu Slow Start

  • Slides: 13
Download presentation
Improving Slow Start In. Kwan Yu

Improving Slow Start In. Kwan Yu

Slow Start? n Slow Start n n n Grows the congestion window size exponentially.

Slow Start? n Slow Start n n n Grows the congestion window size exponentially. Terminates when ssthresh (64 K) is reached or a packet is lost Three cases of Slow Start n n n Initial : when a TCP connection is established Idle : packets are not sent for a few minutes Timeout : transmission timeout

Related Work n Estimating Bandwidth n Hoe’s Approach n n Inter Packet Gap s

Related Work n Estimating Bandwidth n Hoe’s Approach n n Inter Packet Gap s First packet pair to estimate ssthresh RTT / Inter packet gap = cwd (initial congestion window size or ssthresh) Network t

Related Work n Caching n n Source has cache of ssthresh values of destinations

Related Work n Caching n n Source has cache of ssthresh values of destinations When a new connection is from a cached destination, use the ssthresh of the cache.

Related Work n Hu’s Method n n n Congestion : input inter packet gap

Related Work n Hu’s Method n n n Congestion : input inter packet gap is smaller than output packet gap No congestion : input inter packet gap is larger Binary search using about 3 to 4 rounds

Related Work n Hu’s Method

Related Work n Hu’s Method

Goals n Faster Start n n In a large bandwidth*delay path, slow start is

Goals n Faster Start n n In a large bandwidth*delay path, slow start is not fast enough due to TCP’s self clocking Better ssthresh value n Overestimated ssthresh value will cause packet drops

New Methods n History based n With fair TCP, the equal amount of bandwidth

New Methods n History based n With fair TCP, the equal amount of bandwidth will be shared 1 Mb s t 512 Kb

New Methods n History based n When there already k connections, a new connection

New Methods n History based n When there already k connections, a new connection will have B*1/(k+1) share from B total available bandwidth 1/k …… …… s 1/k ? t

New Methods n History based n n Extension to a subnet, inside of which

New Methods n History based n n Extension to a subnet, inside of which delay is negligible and there is no congestion There is a common path from the source to the subnet t Subnet t Host t s Host

New Methods n Estimation based n Weakness of Hu’s method n n The first

New Methods n Estimation based n Weakness of Hu’s method n n The first The binary search method can be misleading with the few errors cwd growth rate is to the 2 nd power (slow) delayed acknowledgement Modification n Regression based method Confidence interval Sequence numbers of packet pair are reversed

New Methods n Estimation based y=x Arriving Inter Packet Gap (ACK gap) Departing Inter

New Methods n Estimation based y=x Arriving Inter Packet Gap (ACK gap) Departing Inter Packet Gap

Reference n n Ningning Hu, P. Steenkiste, Improving TCP startup performance using active measurements:

Reference n n Ningning Hu, P. Steenkiste, Improving TCP startup performance using active measurements: algorithm and evaluation, ICNP 2003. J. Hoe, Improving the start-up behavior of a congestion control scheme for TCP, SIGCOMM 1996.