search

hongzimao / pensieve

Neural Adaptive Video Streaming with Pensieve (SIGCOMM '17)
http://web.mit.edu/pensieve/
MIT License
516 stars 279 forks source link

Missing train_sim_traces & test_sim_traces data at Dropbox #134

Open joybi531 opened 3 years ago

joybi531 commented 3 years ago

Hi, I was trying to run the simulation with the actual data used in the experiment. However, I couldn't find the train_sim_traces and test_sim_traces data at the given link to Dropbox. That might have been removed. Could anyone please provide me the data downloaded from the link that was mentioned in the repository?

hongzimao commented 3 years ago

Thanks for the question! I graduated and lost the access to MIT dropbox. Here are some traces I can dug out from my local machine. Not exactly sure if it's the same as the traces in the dropbox link.

If anyone has a copy of the traces, please let us know. We will upload the traces and use updated link accordingly. Thanks!

cooked_traces.zip

joybi531 commented 3 years ago

@hongzimao Thanks for the help. The traces data that you just shared works fine.

NotSpecial commented 1 year ago

Hi, I am trying to reconstruct the traces used to train Pensieve from the FCC data, but I feel like I am missing a step.

I downloaded the raw data and ran the traces/fcc/load_webget_data.py script. However, this script returns a very different output than the cooked traces you shared above.

It only contains a single throughput line per row, and the throughputs do not match the data in the FCC traces.

For example, these are the first couple of lines for the trace trace_866_http---www.amazon.com:

This is what is returned by the script in this repository:

1358237
1440783
1387260
1383838
1369546
...

And this is what is part of the archive you sent above:

0.0 1.41832
5.0 1.39856
10.0    1.47412
15.0    1.50672
20.0    1.484064

...

Can you help me figure out what I am missing? How can I get from the output of the script in the traces directory to the training data you shared?

Thank you very much in advance!

YuxinLiu2023 commented 1 year ago

Hi, I am trying to reconstruct the traces used to train Pensieve from the FCC data, but I feel like I am missing a step.

I downloaded the raw data and ran the traces/fcc/load_webget_data.py script. However, this script returns a very different output than the cooked traces you shared above.

It only contains a single throughput line per row, and the throughputs do not match the data in the FCC traces.

For example, these are the first couple of lines for the trace trace_866_http---www.amazon.com:

This is what is returned by the script in this repository:

1358237
1440783
1387260
1383838
1369546
...

And this is what is part of the archive you sent above:

0.0   1.41832
5.0   1.39856
10.0  1.47412
15.0  1.50672
20.0  1.484064

...

Can you help me figure out what I am missing? How can I get from the output of the script in the traces directory to the training data you shared?

Thank you very much in advance!

I think he used the functions in plot_cooked_bandwidth.py to calculate the time_all and bandwidth_all.