Force10 Loop-Back -- RBUDP & TCP -- RBUDP Data Size Plots

Data Size Plots

In the following sub sections results are displayed in the form of RBUDP and multiple TCP bandwidth plots as a function of the RBUDP data size. The bandwidth values are presented with average values that are calculated in the areas with only RBUDP, only TCP and with combined RBUDP & TCP traffic types. These flows are generated in the same way as described before in the "Description" subsection. In each plot the following plot traces are presented, entitled:

RBUDP (Only): Contains the RBUDP bandwidth results in the area with only this type of traffic.
UDT (& Iperf): Contains the RBUDP bandwidth results in the area where the RBUDP & TCP traffic types are combined.
Sum Iperf (Only): Contains the sum over the TCP Iperf flows in the area with only this type of traffic.
Sum Iperf (& RBUDP): Contains the sum over the TCP Iperf flows in the area with combined UDT &TCP traffic types.

In the plots of the average bandwidth values from the RBUDP and / or TCP traffic types listed above are presented as a function of the RBUDP data size when one threaded parallel flow per node pair has been used. The figures are displaying the results of tests from 5 × 5 to 11 × 11 node pairs as usual grouped into two VLAN's where each node is sending a TCP flow to all other participating nodes in the opposite VLAN.

RBUDP, Sleep TCP; P = 1, N = 5

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 5 × 5 node pairs, using one parallel flow per node pair.

RBUDP, Sleep TCP; P = 1, N = 6

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 6 × 6 node pairs, using one parallel flow per node pair.

RBUDP, Sleep TCP; P = 1, N = 7

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 7 × 7 node pairs, using one parallel flow per node pair.

RBUDP, Sleep TCP; P = 1, N = 8

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 8 × 8 node pairs, using one parallel flow per node pair.

RBUDP, Sleep TCP; P = 1, N = 9

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 9 × 9 node pairs, using one parallel flow per node pair.

RBUDP, Sleep TCP; P = 1, N = 10

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 10 × 10 node pairs, using one parallel flow per node pair.

RBUDP, Sleep TCP; P = 1, N = 11

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 11 × 11 node pairs, using one parallel flow per node pair.

In the equivalent results are displayed as in the with the exception that there are now five threaded parallel TCP flows per node pair used while in the previous situation only a single TCP flow per node pair had been used with the constraint that the sum overall parallel flows per node pair had been kept the same in both situations.

RBUDP, Sleep TCP; P = 5, N = 5

RBUDP, Sleep TCP; P = 5, N = 6

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 6 × 6 node pairs, using five threaded parallel flows per node pair.

RBUDP, Sleep TCP; P = 5, N = 7

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 7 × 7 node pairs, using five threaded parallel flows per node pair.

RBUDP, Sleep TCP; P = 5, N = 8

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 8 × 8 node pairs, using five threaded parallel flows per node pair.

RBUDP, Sleep TCP; P = 5, N = 9

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 9 × 9 node pairs, using five threaded parallel flows per node pair.

RBUDP, Sleep TCP; P = 5, N = 10

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 10 × 10 node pairs, using five threaded parallel flows per node pair.

RBUDP, Sleep TCP; P = 5, N = 11

. Single flow RBUDP and multiple flows TCP bandwidth as a function of the RBUDP data size. In the left plot the RBUDP flow has been delayed and in the right plot the TCP flows which were defined by all possible 11 × 11 node pairs, using five threaded parallel flows per node pair.

From the results in the following conclusions can be drawn:

There is no significant difference in the results where the delay of the RBUDP flows (left plots) are compared with the delay of the TCP flows (right plots).
When there is only RBUDP type of traffic, variation of the RBUDP data size hardly has any influence. Probably the almost zero round-trip time will be a reason for that.
In all plots their can be seen that the TCP Iperf flows are limiting the RBUDP flow: that is already long before the 1 Gbits/s loop-back interface has been saturated.
With increasing RBUDP data size, the influence of the TCP flows upon the RBUDP flow will become less, resulting at the area with combined traffic in an increasing RBUDP bandwidth and a decreasing TCP bandwidth when the RBUDP. These slopes are getting steeper with increasing number involved TCP node pairs, also with increasing total TCP bandwidth.
However with five parallel flows per node pair the influence of the TCP flows upon the RBUDP flow is stronger (see the previous sub section), resulting in less steep slopes when the same number of TCP node pairs are involved.

RBUDP & TCP Flows

Bandwidth Intv. Plots

Table of Contents

RBUDP & TCP Concl.

F10 Loop-Back Concl.