Network connectivity between SLAC and BINP, Novosibirsk following upgrade

Les Cottrell. Page created: July 26 2002.

Central Computer Access | Computer Networking | Network Group | ICFA-NTF Monitoring

Introduction

The Budker Institute of Nuclear Physics (BINP) in Novosibirsk, Russia is a major High Energy Physics (HEP) site and in particular has strong collaborations with SLAC. In order to facilitae physicsist at BINP collaborating with sites in the U.S. it is critically important for BINP to have good connectivity to ESnet and Internet 2. In December 2000 it was identified that extra capacity was needed and an upgrade was recommended by the technical people (see Network connectivity between SLAC and BINP, Novosibirsk).

After negotiations between Novosibirsk, KEK/Japan, SLAC and DoE funding was made available to upgrade the link, and on March 24, 2002 the link was upgraded.

Upgrade

The link between BINP and KEK was upgraded from a 128kbps to 512kbps land line (fiber optic link). Between KEK and ESnet in the US. The KEK link connects to the ESnet link in Sunnyvale California. The KEK to ESnet link is at 10Mbits/s. The route is shown below:

1cottrell@antonia:~>traceroute rainbow.inp.nsk.su
traceroute to rainbow.inp.nsk.su (193.124.167.29), 30 hops max, 38 byte packets
 1  RTR-GSR-TEST (134.79.243.1)  0.228 ms  0.229 ms  0.201 ms
 2  RTR-DMZ1-GER (134.79.135.15)  0.304 ms  0.294 ms  0.336 ms
 3  SLAC-RT4.ES.NET (192.68.191.146)  0.425 ms  0.409 ms  0.324 ms
 4  snv-pos-slac.es.net (134.55.209.1)  1.335 ms  0.711 ms  0.796 ms
 5  kek-v4-esnet.es.net (198.129.248.41)  129.739 ms  131.302 ms  130.570 ms
 6  kekcis7.kek.jp (130.87.43.7)  132.595 ms  131.808 ms  131.410 ms
 7  192.153.114.138 (192.153.114.138)  237.093 ms  237.117 ms  237.290 ms
 8  rainbow.inp.nsk.su (193.124.167.29)  238.726 ms *  238.604 ms

The round trip time (RTT) and losses on the link for 2002 from SLAC to BINP are shown in the plot below obtained from the SLAC led MICS/DoE funded PingER project. It can be seen that the big variations in RTT (from 250msec to over 800 msec), that are indicative of heavy congestion, disappear after 3/28/02. Also, since the start of 2002, the loss rates have improved from several percent (averaged over an entire month) to less than 0.2 per cent. Losses of several per cent can greatly impede interactive work such as terminal sessions (since the timeouts following a packet loss can make what appears on the screen be several seconds behind what is typed). The derived throughput using the formula
Max(Bandwidth_TCP) ~ MSS / (RTT * sqrt(loss))
where MSS is the maximum segment size (typically 1460 Bytes), indicates that the bandwidth achievable has improved by a factor of 2 from January 2002 to July 2002. Similar improvemnts are seen to other ESnet Labs.

[ Feedback | Reporting Problems ]
Page owner: Les Cottrell