User login

Stephen Eichler's blog

09

Mar

2015

Set up new scamper modifications on a test vantage point. The modes to be used are ICMP per destination limited to seven flow IDs and another mode with many flow IDs. Initially I encountered a bug but now the test run is on the way where both these modes are run for each destination address. This will allow comparison of the ability of these two methods to detect load balancers and compare their efficiencies at doing this task by comparing probe counts.

The churn paper has been updated to include the research on stopping values. It is not yet clear whether the simulation method to estimate stopping values should be included.

02

Mar

2015

20 new PlanetLab nodes where set up for using with the black hole detector. Once this was done a run was initiated.

Load balancers were counted in 6 runs that were previously carried out on PlanetLab. One of these used 128 flow IDs instead of the economical 7. This can probably be used for comparison even though they were carried out sequentially. The comparison is aimed at the number of load balancers found and whether more are found with the fuller analysis.

Changes from the TMA submission were applied to the churn paper. Preparation to apply to a new Journal also has meant reformatting to two columns again.

23

Feb

2015

Results from a corrected run of the black hole in load balancer detector were processed. Some short traces were found that stopped in load balancers i.e. Discontinuities, were found. The detector will now be run on more nodes on PlanetLab to broaden the search.

A scamper routine that uses large numbers of destinations (256) for finding per destination load balancers was programmed. Also a driver to run the old routine and the new routine was programmed. A run of 5000 addresses is now underway on Yoyo. After this test a run will be carried out on my PlanetLab nodes.

More PlanetLab nodes have been activated on my slice. It is amazing how many do not come online when activated. This made it necessary to activate more on two occasions to get the numbers of usable nodes up to the required level of 25.

20

Jan

2015

Further work has been done on the chapter on load balancer prevalence. Statistics on percentage of load balancer nodes out of all nodes were calculated. Analysis of variance was carried out on the data across all vantage points. Categories included counts of all load balancers, counts of diamond divergence points, per destination data and per destination data excluding per flow cases. Another category was per flow or per destination load balancer nodes including internal nodes to load balancers. Not per flow cases of per destination gave a small reduction indicating that these sets are relatively separate already even though per destination load balancers are strictly a case of per flow. Limiting cases to diamond divergence points gave a large reduction.

13

Jan

2015

I started on the load balancer prevalence chapter. This involved adjusting my measures of per destination load balancing to include per flow as is usual. I think that I will still report the per destination load balancers without other per flow load balancers counted as well. Another statistic that I am currently generating is a count of per flow diamond divergence point load balancers. This will be reported as a percentage of uniquely identified nodes found, and will relate well to my per destination load balancer counts as nested load balancers are not found in this case either.

Another run to find black holes in load balancers was started.

09

Jan

2015

Updating and proof reading of several chapters of my thesis have been carried out. These include stopping values, data collection, diamond structure and the many to many scenario.

Some statistics have also been updated including per destination path percentages. Percent per destination load balancers out of all discovered interfaces recalculation is underway.

I have also written up my six monthly report.

16

Dec

2014

Further work has been done on my thesis. I have written a section on suggested usage for the many sources to many destinations scenario. This is based on the cases that we have available which are few to many and the inferred many to few case. The time frame for data collection is based on comparison of published results with my results. This comparison suggests that there is a high rate of temporary change of Internet topology but that there is a slower rate of permanent change. This makes a more workable rate of data collection seem feasible.

09

Dec

2014

The event and time based simulator of Doubletree and Traceroute was extended to cover sources windows of 1, 2, 10 and 100. These all performed well right down to 1. A sources window of one is different to no sources windows because each trace is processed one by one, whereas the later processes them all at the same time. It should also be noted that there is a delay before control information is sent making these two situations more different from each other in terms of total traffic.

Further work was carried out on the related work section of my thesis. This included load balancer turnover and structure as well as Doubletree and black hole analysis.

02

Dec

2014

The IS0 simulations completed and a graph was created. This event based simulator ran a large data set and increased the number of destinations per AS to 20. This made it take much longer to run, as it used much more of the Internet for the analysis. The results make the sources window setting of 500 look like the best option compared to some larger ones, and also to no window.

I added a 'related work' chapter to my thesis and also some information on dynamic routing. In particular I commented on the apparent likelihood of black holes in load balancers. It seems likely that for OSPF IGP networks that they would last a maximum of 40 seconds, as this is how long an adjacent node will wait for a Hello packet before removing the node from its routing table.

25

Nov

2014

Improvements have been made to the introduction and background chapters of my thesis. The section of the introduction on Traceroute is now preceded by more general introductory paragraphs. The background chapter now has a section on Doubletree.

The longer run on the event based Doubletree simulator was partially successful. The Traceroute runs just needed their command lines correcting and a restart of the script. That is now running.

Some of the non event simulator Megatree and Doubletree runs completed. Further runs based on the success of these have been initiated. These are now running.