Google: Difference between revisions
→Networking Followup
Line 110:
;1) Moving of files BW APAC & EMEA
1MBPS speed? whats wrong
Multiple streams
Pearing
Line 136:
;5) Re-Evaluate the design?
Room for improvement?
;Answers
;1)
*Network throughput impacted by TCP window size, Latency and Congestion
*Window Size
Maximum amount of data a sender can send before receiving an acknowledgement. Standard TCP Window Size = 65K bytes
*Bandwidth Delay Product
Amount of data that can be in transit (flight) in the network
Includes data in queues if they contributed to the delay
BDP (bytes) = total_available_bandwidth (bps) x round_trip_time (sec) / 8
Take Packet Captures
Fragmentation
WSF 64kB to 8MB
SACK to minimize data that is resent
Fast re-transmits to reduce delay before resend
*BIC TCP for faster recovery from packet loss
Allows bandwidth probing to be more aggressive initially when the difference from the current window size to the target window size is large, and become less aggressive as the current window size gets closer to the target window size.
A unique feature of the protocol is that its increase function is logarithmic; it reduces its increase rate as the window size gets closer to the saturation point.
*It’s not just about latency, TCP doesn’t like congestion
Adding more traffic produces a negative marginal effect above about 30% utilization
*Window scaling changes the TCP window to:
64KB * 2n (n = window scale factor)
With a window scale factor of 7, which equals a TCP window of 8MB
Single-flow throughput is limited to:
TCP window size / RTT
Without window scaling, TCP is limited to:
64KB / 100ms = 5 Mbps
With CloudBridge default window scale, TCP is limited to:
8MB / 100 ms = 650 Mbps
| |||