IEPM

Status Report for SLAC/DOE IEPM Project

For December 1998 & January 1999

SLAC Home Page

Coordination/Publicity/Deployment

Possible New Sites

We advised the International Atomic Energy Authority on use of the PingER software to monitor links of interest to them and collaborators in Brazil. For a short time it appeared they might come on board and contribute to the HEP monitoring, but we haven't heard from them for a few weeks.

We also received an enquiry from the Greek National Network Center and from an ISP in the Philipines about running PingER.

XIWT

Our work with the Cross Industry Working Group is proceeding. They recently added a new monitoring site ( http://mason.cnri.reston.va.us ). Warren and Les are preparing talks to present to the group early in February.

Attended a voice conference with the XIWT/IPERF group. The main topic was a presentation by NIST on how correlated ping measurements from a to b are with ping measurements from b to a. In the short term seconds to several minutes there seems to be only moderate correlation.

ESnet

We had a voice conference with Becca Nitzan and Mike Collins of ESnet to discuss QoS measurements and preference bit settings. Later had a face to face meeting to go over how to collaborate on DiffServ experiments and measurements.

Followed up on a request from the ESSC International working group to add 4 more Beacon sites to the ESnet/HENP monitoring. Sent email to contacts at the 4 sites requesting a suitable host to monitor. So far, only one of the sites has responded.

In response to a question from a user, we looked at the change in response time between SLAC and KEK following re-routing of the link via the Chicago STAR-TAP. The round trip response time has gone up by between 50 and 70 msec.

NetPredict

Starting to work with NetPredict on a possible phase II SBIR. Followed up on the status of the Technology Transfer proposal Charley Graneieri and Les worked on in the Fall of last year. The Lab directors favor funding this, but it now needs updating since it was written about 6 months ago.

Stanford Computer Science Department

Arranged to give a Stanford Networking Seminar on Internet end-to-end performance in the Gates building on March 4th (see http://netseminar.stanford.edu/). Met with person on campus to coordinate. The abstract is available at http://www.slac.stanford.edu/~cottrell/stanford-cs-abstract.html.

Development/Analysis/Documentation

PingER

Warren prepared documentation regarding the analysis scripts we have developed.

Config Tool

Warren began work on a tool to aid with the configuration of the programs in the PingER package. The tool uses the Perl/Tk perl module. Click here for a screenshot.

ICMP precedence/Trouble-shooting

Last month we put a machine outside the SLAC firewall and behind a router running Cisco test code that allows setting precedence bits. This is to enable tesing the effect of setting precedence headers of ping requests sent to a specific host at CERN. The machine we used for the monitoring turned out to have faulty network chip, so Warren disabled the built in NIC, and installed and configured a new NIC card. Using this we were successfully able to send pings with precedence bits set.

We verified that the precedence bits were set by using a sniffer to look at the packets. Following this we made measurements for about a week to compare ping performance with and without precedence bits set on the path between SLAC and CERN. This path gets utilization up to about 50% but the results showed no impact of setting the precedence bits. Communicated with Denise Heagerty of CERN concerning setting the precedence bits in the reverse direction.

Also working with ESnet to set up an ATM PVC between SLAC & LBNL. We will use this to measure the effects of adding congestion and setting precedence bits on the performance of the link for pings, VoIP and video.

Analzed the ESnet bytes accepted reports to provide long term utilization charts for Walter Toki to present at the ESSC.

Christmas Glitches, BNL down

Around Christmas, several monitoring sites suffered glitches and Warren had to dedicate a lot of his time to finding out the status from the local administrator and politely pushing them to get things back and running. Warren wrote a program checkdata_gif.pl to help us to keep track of gathering data. Fortunately we were able to recover much of the missing data from these glitches, although BNL has been down for much of January due to a security issue.

Network Snapshots

We spoke to ESnet's Michael Collins about anomilies we were observing in routing. He provided us with useful information and Warren adapted his network snapshot program to look at network numbers and AS's rather than just the IP number. Warren has applied the work to produce a snapshot of the network between the XIWT sites.

Validating use of ping

Did some more work on understanding how TCP throughput is related to loss and round trip delay.

Jitter

Developed 2 ways to measure ping "jitter" and compared them to show there is little difference.

Also compared one and two way jitter using Surveyor and PingER data. Sent email to Surveyor folks & Vern Paxson soliciting input. Got a response from Vern.

Tutorial

Updated the Internet Monitoring Tutorial to extend the table of contents, add new sections on the extent of the PingER projects (including the XIWT), the architecture, to provide information on the global and affinity distributions of the ~ 1100 PingER monitor-site and remote-sites, and provide a lot more information on jitter (comparisons of various ways of calculating it, comparing one and two way jitter).

Back to Top

Revised 23 January 1999
URL: http://www-iepm.slac.stanford.edu/about/status/status-01-99.html
Comments to iepm-l@slac.stanford.edu