A bit of background is in order to understand what I’ve been up to.  I’ve been working this week on getting the hang of working with the Planet-Lab infrastructure, and can mostly find my way around it manually now.  I haven’t figured out how to automate the interactions with it in the way that will be needed for this project, but it’s a start.

Planet-Lab is a network of computers around the world which researchers can obtain access to (eventually).  As a user, one gets a “slice”, which as far as I can tell is just a project-specific username.  The user can assign virtual machines on the “nodes”, which are the actual machines.  Users have limited root access on the nodes, and can install software, set up cron jobs (scheduled tasks), and run scripts.

So where has this gotten me? Well, read on….

Things which have gone wrong:

• ran out of space on infotrace.citizenlab.org, the virtual machine I’ve been working on.  Fixable, but not before class.
• had to install a bunch of stuff to get the various planetlab deployment tools and python shell (not) working (see previous point).  Also fixable.
• realized that one can only have one ssh key on planet-lab at a given time.  Fixed this after the usual wait to deploy new keys out onto the infrastructure.
• socket problems on the PL nodes.  Not sure about this one, will need to do more research.
• the Django Book continues to be frustratingly out of date.  Need to bug classmates more.

Things which are working:

• CoDeploy built and running from my desktop, allowing me to deploy files out to the Planet-Lab nodes, run commands on the nodes, install rpm’s, etc
• Adding and deleting nodes from my “slice” via the PL API from a python script.
• TCP traceroutes using tcptraceroute.
• A small python script which allows the nodes to phone home their results with an HTTP POST, but there’s nowhere to POST it to yet…

TODO:

• sort out the strategy for enumerating nodes to be added to the slice, or whether it’s worth just keeping them all added and updated
• script running of queries on the nodes
• choose the options available for running queries
• finish the “results” bit of the django app to receive the result pings from the nodes
• get Steve and Alex working on some of the scripting with me

Architecture

It’s pretty simple.  infotrace.citizenlab.org (DNS isn’t working yet) is running django and a postgreql database; the database could easily be split off to another machine if load requires it.  Queries will be generated by researchers accessing the web interface, and will be sent off to the Planet-Lab “cloud”.  Most of the dirty work there will be done by the various Planet-Lab API’s I’m still trying to get my head around.

User Roles

This should mostly be taken care of with django’s admin module, but I figured it was important to write up as well.

• Administrator: can create leads and researchers, can delete either, can create and set limits on leads and researchers
• Lead researcher (a.k.a. lead): can create / delete researchers up to a certain number, can run queries
• Researcher: can run queries (subject to limits set by researcher)

Requests

Requests which go out to the Planet-Lab infrastructure can take a number of forms:

• one off requests with a single source and single destination
• timed monitoring requests which repeat on an interval,  from a single source to a single destination
• variations on multiple sources, multiple destinations, and monitoring or not
• other types which I’m not thinking of?

Visualization of Output

• kml file (google maps, google earth)
• csv file
• ??

Here’s what I’ve accomplished so far:

• Tracked down U of T’s Principal Investigators for the Planet-Lab network and asked for access for the project
• Set up this blog
• Set up a GitHub account
• Found some similar research
• Read up on BGP
• Explored several tools for doing traceroutes and related network tracing: hping3, nmap’s –traceroute, 0trace, and scapy.

A few links promised to my classmates, which are interesting on their own:

Miles Thibault is working on a business plan for a “Wikimovies” web site.  I think he’d get a lot out of some Long Tail reading:  Chris Anderson’s original article, and Kevin Kelly’s riff on it titled “1,000 True Fans“.

Denis Pankratov and Jennifer Ruttan are working on a really nifty-looking project to do accurate indoor localization with CDMA (that “other” cell phone protocol), and (blah) Ian Goldber’s paper on “Three Protocols for Location Privacy” from last year’s Privacy Enhancing Technologies symposium.

My goals, which were originally for the next two weeks but have been pushed back only one as I’ve fallen a bit behind on the “getting stuff up and running” side of things are:

• Coming up with a database schema for storing connectivity information.
• Getting a basic web interface up and running in django.

I’m working on these first rather than the network underpinnings as we don’t yet have access to the Planet-Lab infrastructure, so the constraints there aren’t entirely clear.  The front-end stuff will likely run on a server at Citizen Lab, so I can get that up and running right away.

-Leigh

I’ll be crossposting some hackerspace-related stuff to the HackLab.To page, and maybe setting up some aggregation for my classmates’ blogs for CSC491.  I post photos over at Flickr but you’ll see some here too sometimes.  Maybe I’ll even write a bit about my other big interests: information security, open source software, and equity / gender in IT issues.

-Leigh