HSN Term Project : Suggestions

[Harish Iyer]

Hi all,

          I along with two other students recently submitted report for our HSN Project.  We wanted to get your feedback on ways to improve or expand the scope. Please find the report attached.

Regards

Harish

---------- Forwarded message ----------
From: Aaditeshwar Seth <as...@cse.iitd.ernet.in>
Date: Sun, Nov 18, 2012 at 7:55 PM
Subject: Re: HSN Term Project Report
To: Harish Iyer <harish...@gmail.com>
Cc: Nitin Rakheja <nitin....@gmail.com>, Chandrika Bhardwaj <chandrika...@gmail.com>


This is quite good. I suggest you quickly try to convert this into a paper. The Mobihoc deadline should be doable: http://www.sigmobile.org/mobihoc/2013/callforpapers.html

Some specific comments:

1. Section 2.1: Mention how many times you did these traceroutes. Important for any experiment you do, else reviewers can question the validity

2. Section 3: What is the subnet mask assigned during DHCP? If it is /16 that will clearly justify your method of plotting only the first two octets. Also check with whois what IP ranges have been assigned to these providers.

3. Section 4: Did you send UDP packets with spoofed addresses? What do you mean by Airtel "terminating the connection"? Describe the experimental setup

4. Section 5.1: Figure labeling is incorrect -- the text says that Reliance gives 10Mbps but the figure indicates MTNL giving 10Mbps. Also, this contradicts with an earlier mention that your USB modem is limited to 7.2Mbps. How is the throughput being measured -- what window is being used for averaging at the receiver? Also, this is download throughput only, so mention that. Do the test for upload throughput as well.

5. Figure 4b -- ping analysis: Y-axis is is ms and not in Mbps. x-axis is what? The RTT values are also better as a CDF. The timeline is fine to give an example, but more useful to show any temporal correlation and hence also measure things like the autocorrelation coefficient for RTTs. This will be sensitive to your ping frequency, so should be done for different frequencies as well.

6. Figure 5 Airtel: This is very interesting. How did you calculate TCP RTTs? I hope delayed Acks did not lead to any mis-counting?

7. Section 5.2: Should be done over multiple days at different ping frequencies. Also log the cell-id and signal strengths to check if a handover happened to cause this drastic change?

8. Section 6: Explain how Netalyzer infers the buffer sizes. Are these stable values? Can you get Netalyzer to report the buffer sizes in bytes as well, or rather, the outstanding data pumped into the network?

9. Section 7: Explain how Netalyzer infers the presence of proxies/caches? DNS Sec -- can you emulate an attack, are there some hacks available through which you can check?

10. Section 8.1: Confirm that TCP RSTs were sent by the server. Also explain hping3 and netem briefly, ie. their capabilities

11. Section 9: What is ows?


For the paper, I think we also circulate this to Vinay and the NICTA folks, to get more feedback. And take help from the folks working in the rural measurements project on the following:
- Since we know that 2G and 3G connections from the same provider share the same path, do they also compete with each other? Can be tested by designing a test with a 2G and a 3G node probing the network at the same time.
- Are the routes taken by packets with and without roaming the same or different?

On 16-11-2012 20:52, Harish Iyer wrote:

Dear Sir,

            Please find the term project report attached.


Regards

--
Harish Iyer
Graduate Student
Indian Institute of Technology, Delhi
India

--
Harish Iyer
Graduate Student
Indian Institute of Technology, Delhi
India