Bulk Thruput Measurements-BNL
Bulk Thruput Measurements | Bulk throughput simulation | Windows vs. streams | Effect of load on RTT and loss | Bulk file transfer measurements
|
Bulk Thruput Measurements-BNL
Bulk Thruput Measurements | Bulk throughput simulation | Windows vs. streams | Effect of load on RTT and loss | Bulk file transfer measurements |
BNL only allows iperf clients, attempts to reach iperf servers at
BNL are blocked at their firewall. Further this site
generally blocks ping and traceroutes.
Another anomaly for this
site is that their Cisco PIX firewall resets the window scaling
option, so the maximum window size is 64 KBytes.
The traceroute
from BNL to SLAC shows the route is via ESnet.
We measured the bottleneck bandwidth to BNL from SLAC with
pipechar.
Since ping was blocked at BNL we measured the RTT from SLAC to BNL using
synack
to be
min/avg/max = 71.789/73.355/178.874 (stdev=10.661)ms
and losses of < 1%.
We measured iperf throughput from rmine601 at BNL
(a Sun with a Gbit Ethernet interface
running Solaris 5.8) to pharlap at SLAC for window sizes of
8, 16 32 and 64 kBytes and streams from 1 to 40.
The results are shown below for July 19, 2001.
The maximum iperf measured
throughput is between 50 and 60 Mbits/s, and this is achieved
for window * streams product of >= 256 KBytes.
Cal Tech | CERN | Colorado | IN2P3 | INFN |Daresbury | Stanford to Daresbury | ANL |BNL | LBL |SLAC To Stanford's Campus
Summary of Bulk Throughput Measurements
Created August 25, 2000, last update August 29, 2001.
Comments to iepm-l@slac.stanford.edu