User login

Stephen Eichler's blog

13

Aug

2013

Monitoring of Caida and Planetlab has been carried out. Completed warts file sets have been downloaded.

The algorithm for finding diamonds has been examined and then used to genereate further statistics.

Work on updating the Internet Simulator to add links to simulate gang probing for doubletree has been carried out.

08

Aug

2013

Churn statistics were generated using counts of load balancers rather than diamonds. These were incorporated into my slides for the six monthly report. This was also to provide a reference point for verifying and improving the algorithm that I use to detect diamonds.

The six monthly report slides were presented.

The next round of scamper data collection on Caida and Planetlab was kicked off and monitored.

23

Jul

2013

Finished downloading warts files for the current planetlab and caida scamper cycle. It looks like I will have to wait a week before starting the next cycle.

Some time was spent finding out why the Internet Simulator does not add extra links to connect all sources with all destinations for doubletree runs. There may be some issues with how to modify the C code involving svn, however for now I have to wait for the crashed scratch disk on Wraith to be replaced.

Further work was carried out on the slides for my six monthly report.

I am also testing yoyo to confirm that it functions correctly on its previous network. The good thing is that I can now access it remotely.

16

Jul

2013

Monitoring scamper analyses and downloading completed cycles has been carried out on Caida and Planetlab.

In an attempt to check my customised scamper program I have needed to carry out a series test runs on the new network connection, as higher packet rates seem to fail to elicit responses when running traceroute. This data will allow Brad to investigate the problem further.

Further development of slides for my six monthly report has been carried out. Some comparison of what we observed from Yoyo with data obtained from Planetlab and Caida has been carried out.

09

Jul

2013

The scamper data collection on Caida was restarted for the next cycle after downloading warts data as it became available from the previous cycle.

I had received notice that my Planetlab slice was to be terminated because the university Planetlab nodes were not online. After the nodes were put back online, I found that my slice was still intact and that I could restart the scamper data collection, just a day or two late.

I generated graphs from the latest Internet Simulator run, even though the many sources and gang probing cases failed. This gave an indication of performance for the remaining 75% of cases. The problem was caused by next hop not found errors, and it was confirmed that a new topology memory map had been produced for the many sources gang probing case. It thus seems that there is something wrong with the process of adding links to make complete paths in this case. Links must be added because the warts data is collected with team rather than gang probing.

I completed my part of the next six monthly report, and I am working on some slides for my supervisors.

25

Jun

2013

Monitored the data collection on Caida and planetlab. Downloaded data as it became ready.

Data was downloaded from Caida (three teams) and from iplane (hubble like) and counts of accumulated interfaces were carried out on the former along with the same on the first run of my analysis data. The idea is to find out what data volume is required to get a reasonable coverage of the internet backbone and to determine if we can simulate it.

The internet simulator was started off in its absolute native state, and with longer abort times in a second run. I determined that the errors from the last run where next hops were not found were associated with simulations with many sources and gang analysis.

18

Jun

2013

Carried out a sanity check on the warts data and detected a bug. Carried out a fix and some testing before sending the fix to CAIDA. This has been implemented and the next round of measurements has been kicked off.

The mostly unchanged run on the Internet Simulator has resulted in some errors. Though the magic number and bus errors are gone, there were still some hash table errors saying that the next hop was not found. I kicked off a completely native run of the Internet Simulator, and after studying the errors from the last run I hope to kick off a run with longer times until stop set control messages abort.

11

Jun

2013

I have been downloading warts data as analysis nodes finish their runs. I have also been working out the most efficient ways to do this, in particular how to avoid two step scp via an intermediate node and designing bash scripts for carrying out a series of downloads.

Some time has been spent on studying the code that drives the Internet Simulator. This has been done with a view to making modified versions to investigate the efficiency of doubletree, and to study the Internet coverage of atlas and hubble in their search for blackholes.

The simulator is currently running and there is now no sign of the error messages that were appearing on previous runs. It is also making us of multiple cores on Wraith which is also a good sign.

I have also been writing up a simple summary on the current overview of this research.

04

Jun

2013

After getting an ssh connection issue fixed, the Caida and planetlab scamper runs were kicked off. Planetlab has one node where scamper does not continue, however there are five Caida nodes where this happens. On further analysis I found some driver cores, so I recompiled the driver on these nodes. I also set debugging on, on one of these for now, so I should get a useful core from this one.

We had a meeting with Tony about the Internet Simulator. I solution to the error messages was found in that the simulator appears to be happy to use incomplete memory maps, if these are accidentally created by aborting the program at a crucial time. The error condition was repeatable so Tony wants a tarball of the damaged scenario, so he can produce a modified version of the simulator where this does not happen. I have run the simulator using the original tarball without any aborted runs, and using my few modifications. This analysis is still running.

28

May

2013

After finding that I could not directly access all of the Caida nodes I was given a Caida node that could access the analysis nodes, however my attempts to log in to this controller node have failed and so I have had to request further assistance in this regard.

The Internet Simulator has produced some new error messages under its modified analysis settings. I went ahead and graphed the results that were collected under these non ideal conditions. I will get a chance to consult with Tony McGregor about these issues tomorrow.