Page 153 - ITU Kaleidoscope 2016
P. 153

ICTs for a Sustainable World




           Total File size = File size in bits (bits) and
           Actual File Transfer Rate = Rate of file transmission in
           bits per second (bps).

                  Transfer  duration  results  of  transferring  843
           Mbytes (6744Mbits) 4K-video file by using three scenarios:
           multi-path  (1:2sec),  using  Path  1  alone  (GIST-B>
           MY>CHULA) and using Path 2 alone (GIST-B>CHULA)
           are shown in Table 6 and Figure 3. The transfer duration
           results are the average results of three times tested with the
           same  parameter  settings.  According  to  the  tested  results,
           file  transfer  duration  decreases  with  upper  bound
           associated  to  the  available  link  bandwidth  capacity  when
           the  target  file  transfer  rate of Tsunami protocol increases
           for all three scenarios. That result trend can be seen when
           comparing the cases of target file transfer rates (100, 200,   Figure 3: File transfer duration over OF@TEIN SDN cloud
           300 and 400Mbps). The required transfer duration of using              playground.
           Path  2  alone  (GIST-B>CHULA)  and  multi-path  (1:2sec)
           are  similar  until  the  target  file  transfer  rate  is  up  to  300   Table 7: Results of average throughput over OF@TEIN
           Mbps when the link bandwidth capacity is enough to carry   Testing Scenarios    Throughput (Mbps)
           the  whole  video  file  traffic.  As  for  using  Path  1  alone   Target Transfer Rate (Mbps)   100  200  300  400  500
           (GIST-B>MY>CHULA), the required file transfer duration   Multi-path(1:2sec)   94.2  188.3  262.9 341.0 394.7
           is  higher  than  the  other  two  scenarios,  although  there  is   Single-path(GIST-B>CHULA)   94.9  188.8  274.9 334.7 354.3
           enough link capacity to carry the whole traffic. The reason   Single-path(GIST-B>MY   91.2  134.9  268.7 226.2 303.6
           is  that  the  longer  round  trip  time  (RTT)  of  125ms  is   >CHULA)
           required for transmitting via Path 1 (via MY route) while
           RTT delay via Path 2 (via a direct link to CHULA) would
           require  105ms.  Moreover,  the  RTT  delay  of  multi-path
           (1:2sec)  is  periodically  switching  between  125ms  and
           105ms. Therefore, the RTT delay is an important factor to
           consider in order to obtain the lower transmission delay.
                  In  the  case  of  target  file  rate  400  and  500Mbps,
           when the link becomes congested, the file transfer duration
           results of our proposed multipath (1:2sec) case outperform
           the  cases  using  Path  1 alone and Path 2 alone. However,
           the  target  file  transfer  rate  of  500Mbps  is  not
           recommended  with  our  multi-path  function  because  it
           requires a longer delay than those of 400Mbps rate. In that
           case,  UDP  packet  losses  increase  due  to  the  overloaded
           links.  The  reason  that  the  file  transfer  duration  of  our
           proposed multi-path (1:2sec) case cannot outperform in the
           cases  of  target  file  transfer  rate  of  100,  200,  300Mbps  is   Figure 4: Throughput over OF@TEIN SDN cloud playground.
           that  using  Tsunami  file-transfer  protocol  limits  the
           maximum transfer rate. Therefore, the transfer rate cannot   Tables  7  and  8  show  the  results  of  file  transfer
           be  more  than  the  specified  target  rate  when  using  our   throughput and actual file transfer  rate  of  varying  target
           multi-path  function.  The  tested  results  of  file  transfer   file  transfer  rate  for  three  scenarios  and  those  results  are
           duration  confirm  that  using  multi-path  splitting  function   average  results  from  tested  three  times  with  same
           achieves  the  lowest  file  transfer  duration  time  when  the   parameter settings. Figures 4 and 5 depict the file transfer
           links  are  congested  and  do  not  have  enough  capacity  to   throughput and actual file transfer rate for three scenarios
           carry the whole traffic by using a single path.    with  various  target  file  transfer  rates.  Those  throughput
                                                              results  confirm  the  remarks  as  discussed  above.  The  file
           Table 6: Results of average file transfer duration over OF@TEIN    transfer throughput shown in Table 7 and Figure 4 are the
                  Testing Scenarios    File Transfer Duration (s)   available  throughput  during  the  file  transferring  period.
           Target Transfer Rate(Mbps)   100  200  300  400  500   According to the file transfer throughput results in the case
           Multi-path(1:2sec)           72.2  36.0  25.7 23.5 26.5   of  400  and  500Mbps  target  file  transfer  rates,  the
           Single-path(GIST-B>CHULA)    71.0  35.7  24.9 30.4 33.0   throughput obtained by using multi-path function is higher
           Single-path(GIST-B>MY>CHULA)  74.3  50.2  25.1 44.3 35.0   than that by using single-path alone.  However, those file





                                                          – 135 –
   148   149   150   151   152   153   154   155   156   157   158