User login

Stephen Eichler's blog




A program was written to calculate a new statistic for measuring load balancer populations. It is the proportion of discovered interfaces that are load balancers. This was coded, run and incorporated into the paper. A plan was also developed for completing the introduction.

Further updates were carried out on the paper, and background reading on the use of the path based load balancer count statistic was also carried out.




Further analysis of Caida run data was carried out to ignore load balancers with a collection TTL less than five when determining population proportions of paths with different load balancer types and different probe types.

The paper was updated with the new results and some other aspects were updated as well.

The six monthly report was filled in with goals and achievements. This may need to be further updated.




The warts analysis program has been upgraded to provide further information for the paper. This has included counting ICMP paths with no route changes and acculating non matching load balancer successor sets for ICMP.

Counts of load balancers at given hop counts have also been calculated for three packet types and three load balancer types. Subsequently to this a program has been written to count percentage paths with one of the three load balancer types where the load balancers with a hop count less than five are not counted. This is to avoid repeatedly counting the same load balancers.

Some of the graphs in the paper have been redone to include ICMP data.

The points raised about the paper at the last meeting have been addresses, though the load balancer path percentages are being run today.




Warts analysis was updated to find matching load balancer successor sets when counting non matching successor sets for each unique load balancing interface. This gave confirmation for the analysis of numbers of successors to the same load balancing interface.

Warts analysis for data from the updated scamper running on Planetlab was designed. This update makes use of more destination addresses for per destination analysis. It would be useful to collect some per destination data with a larger set of addresses.

Changes were made to the section of the paper on turnover. Data with no route changes have been analysed together and incorporated into the text. Data on the numbers of successor sets for each unique load balancing interface were incorporated.

The discussion and conclusions sections of the paper have been written.




Summary 4/12/2013

Updated scamper on Planetlab to use more destination addresses when carrying out per destination MDA (Multipath Detection Algorithm) analysis. The modified version has been run on one node to test the changes.

The code for detecting traces with no change in non load balancer nodes has been applied on a per trace basis to the code for detecting turnover. This has been run to produce more results, and these have been added to the paper.

I have been investigating the validity of the results I have for numbers of load balancing diamonds attached to the same load balancing node. Dumps of unique LB interfaces and successor sets have been examined and an added count of unique successor sets has been added. In calculating this, two sets with one address in common were taken as a match.

More work on the paper has been done. This has included Richards corrections, counts of collapsed load balancers and incorporation of data based on detection on route changes.




Upgraded the method that counts matching successor sets and intersections, to process Planetlab ICMP data.

Wrote a perl method that counts occurrences of load balancers in UDP per packet data and ICMP per destination data. Information from this was included in the paper.

Wrote a c method to process warts data and determine counts of paths where non load balancing remain unchanged. Included results in the paper.

Worked on the diamond analysis section of the paper, and the section on turnover.




A run on Planetlab of ICMP echo probes using MDA at 99.99% confidence has been carried out. These data are to be combined with a previous run to give estimates of stopping values at 99% confidence via CDF analysis.

UDP per flow addresses and ICMP per destination addresses have been collected. These have been compared even though they were collected from different vantage points. A good number have been found in common, helping to explain the different population proportions of different types of UDP and ICMP load balancers.

Continuing with the Internet simulator, a problem seems to have been found with creating control node paths using the large Caida data sets. This seems to fail to make progress and eventually says that the path cannot be created.

Work on a paper for publication has continued. Tables and graphs for the sections on stopping values, load balancer population proportions and methodology have been added along with appropriate text.




A couple of areas that require further checking have arisen as a result of my investigations. These are max widths of diamonds and counts of load balancers based on the set of next hops alone. The latter is based on the results of accumulating load balancer counts over 18 vantage points.

Plots representing the data collected on Caida using three modes of flow ID selection have been generated. This involved running analytical programs to generate the refined data sets.

I started another run on Planetlab to bolster the data that I have collected using MDA with a confidence level of 99.99%.

I gave my talk at the PhD conference.

I was able to do a bit of work on the load balancing paper, namely design and methodology.




Carried out a run to collect high MDA confidence data from planetlab using ICMP probes. Downloaded the warts data began an analysis of probe counts for a given successor count. To do this the analysis software was updated.

Worked on a skeleton of the publication. Bullet points were added to each section suggesting a layout to follow.

Generated a graph of accumulating load balancers as further vantage points are added. This analysis has the limitation that if a destination is repeated at another vantage point that the second occurence will not contribute to the results. Analysis also ran for several weeks to generate the results. It will be necessary to upgrade this first look see to deal with these issues. The idea of this analysis is to get a feel for coverage of the Internet backbone.

Work on the Internet simulator has continued and checking of a memory map file based on Tony's data has been verified as having all necessary connections between sources and destinations, when gang probing is used along with doubletree. Errors had previously resulted when the perl programs that run a set of simulations were used on the same data, so there may be a problem with how the memory map files are shared among simulations. This also suggests that the tracelists should be checked between simulations.

I also contributed to a research proposal of Caida's by providing a note offering to possibly collaborate in the future.




Completed warts data has been downloaded from Caida and planetlab as it became available.

Probe count data has been calculated for different numbers of successor interfaces found. CDF data and a graph has been produced. Native scamper 99% confidence MDA has been found to be a bit short of the disired probability level. It may be necessary to collect some 99.9% confidence scamper MDA data on Planetlab to further investigate this.

Statistics about diamonds from the Caida runs have been collected. It may be necessary to verify the analysis a bit more, and also to analyse the ICMP data from Planetlab.

A couple of problems with the Internet Simulator have been found. One is that during the build phase a case where a new path could not be created occurred. The second is that the simulator produced path not found errors after the build phase in the simulation phase using Tony's data and gang probing, but a single simulation caused no error to occur. This raises questions about the sharing of the memory mapped files between simulations and their completeness before use by other simulations.