Skip to end of metadata
Go to start of metadata

 Increasing the coverage of PingER in developing countries has been difficult so far because it is hard to find hosts which are geographically located within those countries and do not block pings. The usual method of searching for hosts on Google using the top level domain of such countries has proven to be tedious and time consuming task. The various steps involved in this are:

  1. Search for possible hosts on Google using the tld
  2. Ping each host manually to check if it blocks pings
  3. Using  Visual Traceroute or GeoIPtool to confirm whether the host is geographically located in the desired country.

The PingER host searcher is an attempt to solve this problem. It completely automates the above procedure by:

  1. It automatically downloads the results for the required country using its tld from Google. The default number of search results to download is 1000 but it is ocnfigurable and can be specified in multiples of 100 upto a maximum of 1000.
  2. Using regular expressions and pattern matching it searches for hostnames in the results.
  3. After elimination of any duplicate hostnames in the list it starts pinging them individually. At this stage the user can configure the number of pings he wants to send out to each host and the time-out value of the whole ping command. The default value is 10 sec timeout for 10 pings.
  4. After the results of the pings come in, the program filters out hosts which block pings and also those with multiple hostnames for the same IP address, keeping one copy for a single IP address. It also stores the min_rtt for all the hosts in the filtered list.
  5. Finally it checks the hosts in the filtered list on GeoIptool, again using pattern matching to show the top level domain, country, city and lat/long for each host. This information has no guarantee of being absolutely correct but on numerous occasions it was observed that it does possess a high degree of accuracy. The results at this stage can be configured by command line to be filtered optionally by either a threshold min_rtt or the top level domain of the results or both.

The program is available on SVN. You can download it by checking it out of svn using the command.

svn co file:///afs/slac.stanford.edu/g/scs/net/netmon/repo/svn/pinger/trunk/bin/HostSearcher.pl

Criteria for Selecting Hosts

We manually choose sites based on the following criteria:

  • Does the min-RTT make sense compared to sites nearbye.
  • They are really in the country and not a proxy elsewhere (we use GeoIPTools to identify the country).
  • GeoIPTools has a city for the site.
  • The host is a web server - this often enables us to find out more about the site via the web.
  • The site is an educational or government site (in that order of preference).
  • Sites within a country are chosen for diversification, i.e. different cities, different uses (Education, Government, Commercial...), different IP network addresses.
  • We choose sites that appear to have better connectivity (e.g.lower RTT).
  • Where possible >= 2 sites/country (see Hosts per Country per Region).

Examples

Usage

TLD

Date

Country

# Hosts searched

Success

#remaining hosts

Hits

Added

Comment

DJ

10/23/2010

Djibouti

272

No

0

1

0

with webonly, host already included, now pings

CG

10/23/2010

Congo Brazzaville

54

No

0

0

0

even without webonly

CZ

10/23/2010

Czech Republic

550

Yes

0

384

1

Hard to identify universities

GW

11/17/2009

Guinea Bissau

3

No

0

0

0

None matched filter

IR

10/23/2010

Iran

10

Yes

0

2

2

 

MM

10/23/2010

Myanmar

111

No

0

13

0

Several .gov.my nodes all blocked, several nodes from domain 203.81.81

TJ

10/16/2010

Tajikistan

237

Yes

0

8

2

 

CD

8/15/2009

DRC

30

No

4

0

0

 

TD

11/17/2009

Chad

3

No

1

0

0

 

RE

10/20/2010

Reunion

445

No

1

0

0

 

MV

10/23/2010

Maldives

277

Yes

1

3

1

 

IQ

10/24/2010

Iraq

19

No

0

0

0

Also tried search on Iraq universities including uobasrah.edu.iq, www.iraquniversity.net, uobaghdad.edu.iq

ZM

10/30/2010

Zambia

342

Yes

6

13

3

New nodes are outside Lusaka and have satellite

TLD

Date

Country

# Hosts

Success

# Remaining hosts

Hits

Added

Comments

  • No labels