search

nipunarora / parikshan

Parikshan
0 stars 1 forks source link

comments on intro #2

Closed gailkaiser closed 6 years ago

gailkaiser commented 7 years ago

"(Facebook mobile has 2 releases a day, and Flickr has 10 deployment cycles per day [Allspaw J., 2009])." I hadn't noticed before that this claim is from 2009, and almost certainly inaccurate now. Can you find anything more recent about frequent releases?

"This thesis will make the following contributions: First, in Chapter 3 we will present a" Seems odd to say first we'll do blah blah in ch. 3 when you've said nothing about what's in chapter 2.

"Second, in Chapter 5"

You also skip what's in chapter 4, would make more sense to put the study of real world bugs there than as part of ch. 3.

"This allows for debuggers to add a patch/or a fix in the debug-container and check if it fixes the error. Similarly tests can be performed while ensuring the production container output is not changed, and forward progress is ensured. We isolate tests in debug container to ensure forward progress, and no impact on the production-container."

2nd and 3rd sentences repeat the same thing.

"Real-Time Insights: Observing application behavior as the bug presents itself" This assumes you already have the debug container in place before the bug occurs... except you could record/replay the network inputs, but then its not really live debugging.

"No service interruption: Since we are focusing our efforts on service oriented systems, any solution should ensure that there is not impact on the service, and the user facing service should not be interrupted." except you will interrupt during live cloning, so perhaps this should be minimal interruption or unnoticeable interruption

"Although we present a solution that is designed to be general purpose and applicable to a variety of applications, in this thesis we specifically limit our scope to the following:" its not really general purpose, since we cannot claim network inputs are sufficient for other domains. (and the next para essentially says this)

"this technique will provide real-time insights into running systems, and allow developers to debug applications without fearing crashes in the production application" not sure what this means, like using the debugger could crash the application? i'd think the slowdown would be of more concern, plus obviously you cannot singlestep, breakpoint, etc.

"The chapter also includes a case study of 16 real-world bugs, and a survey of 150+ bugs real-world bugs which we use to show that network replay is enough for triggering most bugs in service-oriented applications." earlier you had a number >200, while that is indeed also >150, seems these numbers should match.

the outline (1.8) still leaves the contents of chapter 4 as a mystery...

"Next we look at active debugging which allows debuggers to apply fixes, modify running binaries and verify the bug fix on a parallel execution of an unmodified application." if this is a form of mutable replay you should say so.

"Appendix" ???

nipunarora commented 7 years ago

1."(Facebook mobile has 2 releases a day, and Flickr has 10 deployment cycles per day [Allspaw J., 2009])." I hadn't noticed before that this claim is from 2009, and almost certainly inaccurate now. Can you find anything more recent about frequent releases?

  1. First, in Chapter 3 we will present a" Seems odd to say first we'll do blah blah in ch. 3 when you've said nothing about what's in chapter 2.

You also skip what's in chapter 4, would make more sense to put the study of real world bugs there than as part of ch. 3.

  1. "The chapter also includes a case study of 16 real-world bugs, and a survey of 150+ bugs real-world bugs which we use to show that network replay is enough for triggering most bugs in service-oriented applications." earlier you had a number >200, while that is indeed also >150, seems these numbers should match.

    • The exact number is 217. I will fix this in all the locations.