Final ReportExchange VU-CAIDA
international exchange of scholars for software projects
Partipants
Before I go onto the summary, I first would like to thank the people that made the exchange between the Vrije Universiteit, Amsterdam, The Netherlands and caida (cooperative association for internet data analyses, San Diego, USA) possible. I would like to thank kc claffy, Evi Nemeth, Frances Brazier and Maarten van Steen for their efforts to allow me to do this exchange and Rex (USENIX/NLnet endeavour) for providing everything else needed for the exchange.
CAIDA
Caida is a organization that by means of their skitter tool are actively probing the Internet in order to analyze topology and performance. Skitter measures forward IP paths by recording each hop to many destinations. Using ICMP echo requests, skitter also collects Round Trip Time (RTT) to those destinations. The data collected provides indications of low-frequency persistent routing changes and correlations between RTT and time of day may reveal a change in either forward or reverse path routing. The skitter data can also be used to visualize the directed graph from a source to much of the Internet with their Otter tool. Otter can handle visualization tasks for a wide variety of Internet data, including data sets on topology, workload, performance, and routing. Caida has written many more tools to collect, analyze and visualize data, for example CoralReef and Walrus.
VU/GLOBE project
The GLOBE project of the VU is a novel scalable infrastructure for a massive worldwide distributed system. I worked on GLOBE by writing an application called Loc8 which is able to locate users over the internet using the GLOBE system.
Results
The results of the exchange are that the VU now has a GLOBE service running in San Diego at caida and we will install a skitter machine at the VU to help caida do their analyses on the internet. The GLOBE machine in San Diego will give us more insight in the behaviours at long distances and will provide a better real world test case. The skitter machines at the VU will provide caida with data of the internet as seen from Amsterdam. As GLOBE matures we can use caida's data to find the best locations for our GLOBE boxes to provide a good service coverage and their data on bandwidth and round trip time etc. will help us to optimize GLOBE.
Future
Future projects at caida include the integration of bandwith estimation data into a database with Otter and/or Walrus as a front end. The data will be collected by a tool suite called netchar which uses features from Bruce Mah's pchar, Allen Downey's clink, and Constantinos Dovrolis' pathrate. The data processing back end refines and extends capabilities in Allen Downey's clink, and will assemble a view of the subnets connecting the hosts and estimates the bandwidth and latency of each link appearing in a path. Finally, the graphical user interface will layout the graph generated by the back end and allow the user to drill down into the collected data and estimated characteristics.
Working at caida
The commute to work took about 40 minutes of which 20 minutes walking and 20 minutes by bus. Riding the bus was free because students could ride free in a limited area around the UCSD campus. Caida has no strict hours so that gave me the abillity to avoid rush hours by going in later and leaving later. The office I worked in I shared with many co-workers (called elves). Some of them would already be there when I came in and some came in later. The office (called the elf/bin) is also shared with a couple of students working for caida and they came in whenever their classes allowed them to. Because the other co-workers worked in different offices, instead of having to walk to their offices, we used a communication system called the 'moo'. The moo is like a chat-room application on which you can type your questions to people or announce other things.
We had meetings twice a week. One general meeting where everyone gets a chance to explain what they are working on at the moment and important things that everyone needs to know. There is also one skitter meeting only for people using/working with skitter where topics related to skitter are discussed. Usually the meetings are under one hour but sometimes they can take longer.
When I had problems with hardware/software the sysadmin guys fixed it as fast as they could so I wouldn't be stuck on it for too long. When I ran into problems during an assignment, help and guidance was always available.
After office hours, caida sometimes threw a party either at the beach or in someone's house, which was very nice. Caida puts a lot of effort in making me and other co-workers feel at home and it creates a good working atmosphere. Co-workers also took me to off-campus lunch and sometimes gave me a ride home.