Mocking with Tensorflow APIs

[trisongz]

Hi, I ran across this project and after looking through the issues, I wasn't able to find any other comments related to this, so I had a few questions.

My particular use case would be to test and see whether specific Tensorflow APIs would detect the endpoint as a GCS object.

• Would you happen to know whether the mock endpoint is recognized in Google Cloud's logging as GCS?
• Assuming it is recognized, what would be some factors to consider that would affect performance besides network bandwidth, disk I/O, CPU and RAM, for doing large batched file reads (20x100MB every 30 seconds)?

If there aren't currently answers to the above, then I'll probably look to figure it out and report back, but if there are, it would be helpful as a starting point. Thanks!

[fsouza]

Hey @trisongz, thanks for opening this issue. Unfortunately I do not know the answer to those questions, but I assume the logging format would not be recognized, since we're having the default log format provided by gorilla/handlers and I imagine Tensorflow would expect some form of structured log.

Feel free to dig into this and let me know if I can help in any way.