Blog Post Overview
All submissions for this blog, either for guest contributors, or for Stanford CS244 Spring 2012 students, should be a 2-4 page blog post that describes your project, focused on a single question with a single result, but with enough backing figures, data, and explanation to convince us that you have done an experiment with trustworthy results. We want you to:
- Replicate an existing result (good), show a limitation of CS244 Mininet that prevents you from replicating the result (negative results are equally good), or produce a new result (even better)
- Implement experiments in a way that is easily replicated by others – ideally, such that anyone can run a single script that runs the experiment and generate the output graphics.
- Convince us that your experiments are valid.
- Explore new and illustrative avenues; for example, by studying the sensitivity of your result to its input, such as network topology, traffic parameters, line rates, and queue sizes.
The template below covers pieces that we expect to see in your blog post. Each piece should definitely be covered in the post, but you have the creative freedom to tweak the template if it improves the story in your post.
Blog Post Template
Key Result(s): one-sentence, easily accessible description of each result.
Source(s): papers (appropriately referenced, and with a link), websites, videos.
Contacts: names / emails.
Introduction: Briefly introduce the paper, its motivation, and explain the graph you hoped to replicate.
Methods: How was the experiment implemented and run? What software and algorithms did you use? This section might include graphs to validate that a project works as expected.
Results: Describe your results. Do they appear to be the same as the paper? Assuming you could replicate the key results, what other results do you have?
Lessons Learned: What did you learn? Explain scaling limits of your experiment, any aspects of the paper that you found to be underspecified, stuff in Linux you needed to fix, and anything that would not be obvious or apparent after reading the paper. What were your implementation experiences – what was hard, and what just worked? Could you use Mininet HiFi as-is, or did it require changes? Did running EC2 present any issues?
Instructions to Replicate This Experiment: Provide sufficiently detailed instructions for anyone to repeat the results, ideally in less than 15 minutes and with a single script. Note the location of all code, along with specific version numbers (like SHA1’s for git). A custom AMI would ideally not be needed to replicate the experiment, but if this makes replication much easier, then go ahead. We prefer if a single Ubuntu 12.04 EC2 c1.xlarge can run everything, even if you need to update the kernel with custom .deb packages.