Skip to end of metadata
Go to start of metadata

 

Purpose

PingER (Ping End-to-end Reporting) is the name given to the Internet End-to-end Performance Measurement (IEPM) project to monitor end-to-end performance of Internet links. The project now involves hundreds of sites in many countries all over the world and we are actively seeking new partners for this project.

Examples of Use

See here.

Case Studies

PingER Validation

 

PingER Analysis

PingER Operations

PingER Deployment

PingER Databases

Report on converting PingER flat files to a SQL database, Ghulam Nabi, M. Farhan Maqsoos, Bilal Naseer.

PingER Data

Toolbox

Porting PingER Archive to NUST

PingER2

Installation

http://www.rnoc.gatech.edu/pinger2/, also Johari has created a shell script to automate the installation of pinger package in Ubuntu/Fedora/Centos Linux distro.  The finalized installation script for pinger software for CentOS/Fedora based distribution is available from the following url (past the step by step tutorial): http://pinger.unimas.my/pinger/install-tutorial.php. Kashif assisted in testing the script.

SLAC

http://confluence.slac.stanford.edu/display/IEPM/PingER2+at+SLAC

Raja Asad Khan, October 2012

PingER2 Data Flow & Access

Work in Progress

  1. Automating the creation of offsite.nodes directly from nodedetails.
  2. Automating the creation of beacons.txt directly from nodedetails.
  3. Document procedures for editing existing nodes.

ePingER

This is a project to provide low cost, low power PingER monitoring or remote stations.

PingER Futures

TULIP

Tulip Web Site
Old TULIP Web site
Tulip-CBG project report, by Ghulam Nabi, M. Farhan, Maqsood, Buialal Naseer, 2011
TULIP Analysis of alpha as a function of RTT and Region for Europe, North America and Pakistan, by Raja Asad Khan, October 2012.
Discussion on using International gateways to provide alpha between regions.

Interesting papers

Calculating the intersection area of multiple circles

Visual Traceroute (VTrace)

Introduction

Visual Traceroute is a new feature that can be accessed by calling reflector.cgi with function=vtrace. Currently it can provide visual traceroute from 29 landmarks (2 in USA and 1 per each of 27 countries). Some of these are listed below, for a complete list check the from dropdown:

  1. www-wanmon.slac.stanford.edu, SLAC National Accelerator Lab, near San Francisco, California, USA
  2. pinger.cern.ch, CERN, Geneva Switzerland
  3. maggie1.seecs.edu.pk, National University of Science and Technology, Islamabad, Pakistan
  4. mel-a-ext1.aarnet.net.au, Australia's Academic and Research Network, Melbourne, Australia
  5. pinger.fsktm.um.edu.my, University of Malaya, Kuala Lumpur, Malaysia
  6. lblnet-test.lbl.gov(131.243.24.12), LBL, Berkeley, California, USA

It works by first finding the traceroute to the target using the traceroute.pl installed at the landmark node and then geolocating each of the intermediate hop routers using TULIP. It can be accessed from http://tulip.slac.stanford.edu or here.

We use TULIP's dynamic ping-based geolocation as compared to say database methods such as used by MaxMind derived tools like GeoIPTools, since often router locations in the database tend to be given as at the corporate HQ that owns the routers (e.g. ESnet routers may supposedly be located in Berkeley).

Performance

Normally it would take about 10 mins to perform a complete visual traceroute but we have incorporated caching to speed up the process. As the router locations are found they are automatically cached in /nfs/slac/g/net/pinger/tulip/cachetr/cache.txt

If the location of all the routers in the requested vtrace are cached then it takes only about 10 secs to give the output.

Working Specifics

A subset of the Maxmind city database is used to find the nearest city to each router, only cities with population greater than 100,000 are considered.

  • The database used is /nfs/slac/g/net/pinger/tulip/citidb.txt

Both PingER and PerfSONAR landmarks can be used as the traceroute source

Visual Traceroute can be called directly from URL. The format is:

The From landmarks drop down list is populated from /nfs/slac/g/net/pinger/tulip/from.txt .This list if checked and updated daily from crontab using vtracefromchk.pl

Certain routers can not be geolocated because they respond to only a few pings or don't respond to them at all. A script vtrace0chk.pl tries to geolocate such routers in the cache and updates the cache accordingly each night.

Both vtracefromchk.pl and vtrace0chk.pl are placed in /afs/slac/package/pinger/tulip/ and executed via trscrontab in pinger.slac.stanford.edu for user pinger.

Design

The tabular traceroute is shown together with the Autonomous System Number (ASN), if known, the router's location coordinates and distance between hops. Clicking on the IP address will take you to the TULIP geolocation utility. Clicking on the ASN will provide information on the ASN. 

Two traceroute paths are shown on the google map. One (shown in red) is drawn using TULIP and the other (shown in blue) is drawn using MaxMind (an IP host location database). The hops are shown as appropriate numbered markers, the error in the estimated location is shown as a red circle. See the figure below:

This figure shows the traceroute from SLAC to NUST Pakistan. http://www-wanmon.slac.stanford.edu/cgi-wrap/reflector.cgi?function=vtrace&target=nust.edu.pk

Comparisons between MaxMind and TULIP

This presentation was sent to Harvey.

University of Malaya, Kuala Lumpur to NUST IslamabadNUST Islamabad to University of MalayaCERN to NUSTMelbourne to NUSTMelbourne to CERNMelbourne to SLACCERN to UCSD
SLAC to CERNSLAC to ICTP TriesteSLAC to NUSTSLAC to UCSDSLAC to MelbourneSLAC to Waikato New ZealandSLAC to U Oregon
SLAC to AndoraSLAC to MalaysiaDallas to DallasDallas to Dallas detailSLAC to emergency.slac.stanford.eduDetail on emergency 
.

GeoIPtools ends in Brazil.
VTrace correctly shows the route to the backup site in Illinois

 
   
Strasbourg to UMLBL (Berkeley CA) to UMSLAC to Cape TownAlgiers to SLACBolivia to BNL New YorkValparaiso Chile to SLACMumbai to NUST, Islamabad
Beijing to NUSTBangladesh to SLACBangladesh to CERNBangladesh to Rio de JaneiroBurkina Faso to NUSTBurkina Faso to SLACBurkina Faso to KEK Tokyo
Taiwan to NUSTEngland to Washington DCTULIP location of Sofia UniversityRio to Tokyo
 

The England to Washington DC is interesting since VTrace, Maxmind and undns (orange line) give very different results for the target. Despite Sofia University advertising itself as being in Palo Alto the web server is actually near Washington DC. Also the route does not go from London to Paris as indicated by Maxmind.

Complementarity of VTrace and MaxMind

SLAC to Portugal
SLAC to Braga Portugal overviewSLAC to Braga, W. Coast USSLAC to Braga via Netherlands and UKSLAC to Braga, Portugal detailTracerouteTULIP geolocation of Braga

By providing both VTrace (red) and Maxmind (blue)  routes, the results can be complementary in ascertaining the correct route. This is illustrated in the plots below from SLAC to a host in Braga Portugal (193.136.19.13). The left hand map shows how  VTrace accurately traces the route across the US, while Maxmind jumps from SLAC to the middle of the US, back to Berkeley and thence to Europe. This is shown in more detail in the W. Coast US map. Moving futher right, Vtrace shows the route through the Netherlands and the UK, while Maxmind  misses the Netherlands altogether. The Portugal detail indicates that Maxmind is more accurate in showing the final route to Braga. However, note that the Vtrace accuracy circles indicate the considerable uncertainty in the location of the last hop, hence improving one's confidence.  The traceroute figure which indicates that hops 12 and 13 are likely to be in the Netherlands and the UK given their naming conventions (note the top level domains .nl and .uk in the names). The TULIP plot shows how the  paucity of nearby landmarks (black market balloons) results in the poor geolocation of the Braga site.

Cape Town to Tokyo
Cape Town to TokyoAmsterdam hop 9 seen by TULIP

 

Maxmind catches hop (8) in Europe (but gets hops 8 and 9 in London rather than Amsterdam) that VTRace (hop 8 does not respond to pings) misses and undns (orange line) misses both hops 8 and 9. VTRace catches the Johannesburg hop (4), and Pretoria hops (5) that undns and Maxmind miss. VTRace and undns catch the New York hop (10) that Maxmind misses.

SLAC to Burkina Faso
SLAC to Burkina Faso

Maxmind gets the location of the hops in France, but does not do well in the US.

Pakistan to Washington State
Pakistan to Microsoft WA

For the Pakistan to hotmail.com (Microsoft WA) undns gets off track and goes to Australia.  VTrace correctly gets hop 6 as being in Karachi (while Maxmind puts it in the middle of Pakistan), however VTrace  loses track after getting to Frankfurt Germany (since hops 9-13 do not respond to pings), while Maxmind gets the end destination as being Microsoft's home of Redmond Washington. 

SLAC to Sudan
SLAC to SudanLocation of 16th hop
 

VTRace gets the route across the US and to France, but not the last hops to Sudan. DeoIPTools gets Sudan as the final destination.

Melbourne to French Polynesia
Melbourne to French PolynesiaDetail for AustraliaDetail of VTrace & undns at MelbourneDetail of W. Coast USDetail showing VTrace & undns at SeattleMalaysia to French PolynesiaJapan to French Polynesia
    

Maxmind gets the location of the host in French Polynesia. However the route in Australia and the US is badly off. Vtrace and undns agree for Australia and for hops in the US. VTrace does not work for French Polynesia since the underlying TULIP does not attempt to locate a host where the nearest Tier 0 landmark (SLAC) has > 60msec min_RTT. N.b. Australia's route to French Polynesia goes via the U.S. as does Malaysia's, and Japan's. The Japan traceoute is interesting in that undns shows the route as going via Chile. THis requires more investigation. The landmarks in China and Bolivia were unable to make traceroutes.

Miscellaneous

The traceroute from SLAC to UTM, Johor Bahru, Malaysia for MaxMind goes to central US, Japan, then via China and terminates in Sarawak. The VTrace goes from SLAC to Japan (undns confirms this), then to Kuala Lumpur Malaysia, then to Johor Bahru. It then takes an excursion to the Gulf of Thailand (hop 14) and appears to stop there (middle image below). On more detailed inspection (right hand image below), it is seen that it then returns to Johor Bahru and UTM (hops 15-16).

Global view of SLAC to UTM, Johor Bahru,MalaysiaE. Asia part of routeLast few hops
  

Looking in more detail at hop 14, only 4 landmarks have a minimum RTT of < 60ms. Three are in Thailand and one in Kuching Malaysia. See left image below. There are other landmarks in the area, one at UTM in Johor Bahra, Malaysia, another at UM in Kuala Lumpur. However, for unknown reasons neither UTM nor UM were able to ping 161.139.244.5.

Landmarks with < 60msec minimum RTT to 161.139.244.5Geolocation of 161.139.244.5 
 

 

Other Visual Traceroutes

  • CQ Traceoute, one server in St Paul, Minnesota, gives country, lat/long and distance of router from server, shows map as static image
  • DNSTools, one server in Frankfurt, no lat/long, provides map 
  • WhatismyIPaddress, only takes IP addresses, tabular output identifying city, map shows end point only
  • Montis, has 3 servers, in US (Pacific Northwest), Europe (Ireland) and Asia/Pacific (Singapore), provides map, no lat/longs or cities
  • Visualware, can choose server region and a limited number of target regions, requires Java, does not support traceroutes
  • VisualRoute, buy application, identifies city,  no lat/longs. no map
  • Free Visual, download free application, don't see a map
  • GTrace, download free application, has map
  • VTrace multiple servers, gives lat/long plus error, provides map,city, country code

Of these Montis appears to be the most reasonable to compare with.

Montis Server in Singapore to UM Kuala Lumpur  
  
  

PingER LOD 

This is a project to provide Linked Open Data access to PingER data. 

PingER Warehouse

PingER and BigData

Leveraging PingER big data via  modified pingtable.

PingER and perfSONAR

At one time perfSONAR supported PingER. This was discontinued October 2014.

Comparison of PingER and perfSONAR.

Future PingER Projects

Progress Reports

PingER Papers and Presentations

Pinger Meeting notes

Page: Anomlaous Hosts Page: CERN View Page: TULIP Circle selection for Apollonius results Page: Conditional Port Probability Page: Database Synchronization Page: Development Indices and PingER Correlations Page: ePingER Project Zambia Page: Monitored Nodes Status Page: MOS Page: PingER2 at SLAC Page: PingER2 Data Flow Page: PingER Beacon Expansion Page: PingER Case Studies Page: PingER Daily Management Page: Pinger Data Analysis Page: PingER data flow at SLAC Page: PingER Data Processing Workflow Page: Pinger-DB Implementation details Page: pinger-db usage Page: Pinger Deployment Page: PingER Host Searcher Page: PingER Meta Database Page: PingER Metrics Intensity Maps Page: PingER Metrics Motion Chart Page: PingER Monitor node format Page: PingER NODEDETAILS Page: PingER Regions Page: PingER Route Visualizer using Traceroutes Page: PingER Smokeping Integration Page: Possible Remedies to Promote the Internet in Developing Regions Page: Shahryar Progress Report Page: Usage Information Page: Problem Cases Page: Web Interface for the PingER Meta Database Page: Deploying table.pl Page: Afghanistan Case Study Page: Future PingER Projects Page: Progress Reports Page: Updating ping_data.pl and traceroute.pl Page: Implementation of Relational archive site for PingER Page: Derivation of Directness of Connection Page: Mapping of files Page: PingER Data Explorer Page: The Arab Spring 2011 and Internet instability seen by PingER Page: Installation Overview Page: Pinger+PerfSonar schema Page: Status of Pakistani monitoring hosts Page: Traceroute Archive Page: UETTAXILA Analysis of changed Network. Page: Comparison of PoP (pingerjms.pern.edu.pk) and Non-PoP (pinger.usindh.edu.pk) Page: Happy Planet Index (HPI) Page: TULIP-List of MATLAB tool Boxes Required Page: Comprehensive Nuclear-Test-Ban Treaty Organization Page: Duplicate packets Page: Pinger Schema Queries Page: New PingER developers Page: Publicity Page: Economic Indexes Page: ePingER project Malaysia Page: Choice of packet sizes Page: History of growth in PingER hosts this millennium Page: Make tarball for pinger-2.0.x Page: Accessing PingER data Page: Finding PingER Hosts Page: PingER Linked Open Data for Semantic web access to analyzed data Page: PingER install at Daffodil university Page: Copy of 20140305 SLAC SEECS and UNIMAS Meeting Page: Using git to prepare a paper with others Page: PingERLOD publications etc. Page: Leveraging PingER big data with a modified pingtable for event correlation and Clustering Page: Events relevant to PingER Page: PingER and perfSONAR comparison Page: PingER on a Virtual Machine at SLAC Page: ePinger Project at SLAC Page: ePingER on Android phone Page: PingER Data Warehouse using Big Data with Cloudera on Nebula Page: Space recovery for /nfs/slac/g/net/pinger/ reclaiming Page: Examples of the use of PingER Page: Atlantic magazine interested in uneven development of the Internet Page: Visualizing PingER data Page: Blocking by firewall Page: Tulip reflex calling sequence Page: PingER Futures Page: PingER deployment at Guangdong University China Page: PingER at DESY Page: SLAC PingER MA administration Page: Ping Correlation with FTP Page: PingER Script vulnerabilites Page: Add new metrics to PingER Page: Correlation of PingER throughput and Fertility Page: PingER Papers and Presentations