Bulk Thruput Measurements-SLAC TO Stanford's Campus

Bulk Thruput Measurements | Bulk throughput simulation | Windows vs. streams | Effect of load on RTT and loss | Bulk file transfer measurements

SLAC to Stanford campus

The first plot shows the thruput from SLAC (pharlap.slac.stanford.edu) to Stanford (elaine1.slac.stanford.edu) measured on Wednesday September 20, 2000. The second plot shows the same measurement made on Saturday October 28, 2000 this time to elaine28.stanford.edu. The the third plot shows the thruput from Stanford campus (elaine17.stanford.edu) to SLAC (pharlap.slac.stanford.edu a Sun 4u Enterprise 4500 running Solaris 5.6) measured on Saturday October 28, 2000.

The elaine hosts at Stanford are a cluster of Sun 4u Sparc Ultra-60s running Solaris 5.7 and providing interactive services to many users. As such they are typically heavily loaded during normal hours.

Between September 20 and October 28, the link from Stanford to SLAC was upgraded from OC3 (155Mbps) to OC12 (622Mbps), and the SLAC interface to the outside was changed from a 100 Mbps half duplex FDDI ring to a gigabit ethernet. However, the major limit on performance is probably the 100Mbit/s fast Ethernet connections to the end nodes. The path characteristics measured by pchar indicates that the bottleneck is at the link to pharlap.

At one time we had the pharlap host cofigured for 100 Mbps half duplex but the switch port it was connected to configured for 100Mbps full duplex. In this case the maximum thruput from SLAC dropped to 9 Mbits/s or almost a factor of 10 drop in performance. At the same time pchar showed no apparent loss in bandwidth.
SLAC to Stanford bulk thruput

Sample Bulk Throughput Measurements

BACK TO MAIN PAGE

Summary of Bulk Throughput Measurements

Created August 25, 2000, last update August 29, 2001.
Comments to iepm-l@slac.stanford.edu