<?xml version="1.0" standalone="no"?>
                    <!DOCTYPE div SYSTEM "/www/backend/www-xml-443/dtd/caidaML.dtd">
                    <!-- do NOT ERASE the DOCTYPE declaration! --><div>


<tr bgcolor="#f4f4f4">
  <td>
<font face="helvetica,arial" size="2">
<b>URL:</b>
</font>
</td>
  <td>
<font face="helvetica,arial" size="2">
<a href="http://www.ripe.net/ris/Talks/0211_IEPG/">http://www.ripe.net/ris/Talks/0211_IEPG/</a><br/>
<a href="http://www.ripe.net/ris/beacon.html">http://www.ripe.net/ris/beacon.html</a>
</font>
  </td>
</tr>


<tr bgcolor="#e9e9e9">
  <td>
<font face="helvetica,arial" size="2">
<b>Entry Date:</b>
</font>
</td>
  <td>
<font face="helvetica,arial" size="2">
2003-05-14


</font>
  </td>
</tr>


<tr bgcolor="#f4f4f4">
  <td>
<font face="helvetica,arial" size="2">
<b>Background:</b>
</font>
</td>
  <td>
<font face="helvetica,arial" size="2">
A BGP beacon is an unused BGP prefix (unused in the sense that no traffic
is ever forwarded to the prefix) that is intentionally announced and
withdrawn periodically at known times.  Announcements and withdrawals
are usually spaced apart by two hours.  Beacons are useful for studying
BGP convergence, and the concept is related to Craig Labovitz's
technique of injecting faults into the interdomain routing system
as a means of studying convergence.


</font>
  </td>
</tr>


<tr bgcolor="#e9e9e9">
  <td>
<font face="helvetica,arial" size="2">
<b>Datasets:</b>
</font>
</td>
  <td>
<font face="helvetica,arial" size="2">
<ul>
<li>RIPE RIS BGP Beacons: one /24 prefix at each of 9 RIS route collectors</li>
<li>BGP updates in Oct 2002 caused by beacons; observed at RIS (?)</li>
</ul>


</font>
  </td>
</tr>


<tr bgcolor="#f4f4f4">
  <td>
<font face="helvetica,arial" size="2">
<b>Results:</b>
</font>
</td>
  <td>
<font face="helvetica,arial" size="2">
<ul>
<li>statistics on number of triggered BGP updates per event per peer:
  <ul>
    <li>observe more messages for withdrawals than for announcements</li>
    <li>most announcements cause fewer than 5 update messages</li>
    <li>most withdrawals cause fewer than 10 update messages</li>
  </ul></li>
<li>statistics on latency of first BGP update per event per peer;
    distributions are similar for announcements and withdrawals;
    centered at 30 seconds</li>
<li>statistics on latency of last BGP update per event per peer:
  <ul>
    <li>latency is higher for withdrawals than for announcements</li>
    <li>latency for most announcements is under 100 seconds</li>
    <li>latency for most withdrawals is under 300 seconds</li>
  </ul></li>
</ul>




</font>
  </td>
</tr>
</div>

