search

gaul / are-we-consistent-yet

Compare eventual consistency of object stores
Apache License 2.0
169 stars 12 forks source link

1 byte payload? #9

Closed jacktasia closed 9 years ago

jacktasia commented 9 years ago

Thanks for making this. It's super interesting. Although it appears you're using one byte payloads...is this correct? If so, it seems like that makes this merely a theoretical exercise since in "real world" usage payloads would almost always be way, way bigger. I would think the larger the payload the greater the chance for there to be consistency issues. Unless I am missing something it seems like ideally you'd have multiple tables in your README using different payload sizes. Thoughts?

gaul commented 9 years ago

@jacktasia Thank you for your feedback; AWCY should allow a configurable payload size. Using a larger payload could expose more eventually-consistent behavior, for example, when a object store uses remote replication like AWS us-standard using both N. Virginia and Pacific Northwest. Would you like to try adding this feature?

jacktasia commented 9 years ago

@andrewgaul Thanks for the response. Your us-standard example is exactly what I was thinking about. My main concern is people will use the data in the README as evidence that consistency issues are not actually real, which I think is misleading with the current payload size. While I would like to add in configurable payload sizes as a PR I won't have time any time soon, but I do think in the short term the payload size should be explicitly mentioned in the README table so people are fully informed if they are making decisions based off this information. Unfortunately I've found people will very rarely bother to read the code.

I'd be happy to do a README PR if you agree.