Question about CacheLib's NVM concurrency

[wangyuyue]

I notice that in both allocator/nvmcache/NvmCache-inl.h and navy/scheduler/ directories, there exists some code related to the handling of concurrent NVM operations for objects of the same key. So what is the difference of responsibilities between them? allocator/nvmcache/NvmCache-inl.h is quite complicated, and I really hope somebody can solve my confusion. Thanks a lot.

[haowu14]

Hi sorry for the late reply.

If I were to be very specific about the wording:
NvmCache-inl.h is more on concurrency control: making sure that different threads will be able to see a consistent view. For example, lookup would abort on-going puts, etc.

navy/scheduler/ is about parallelism: Given the incoming NVM jobs, how to put them in multiple queues and execute them in parallel.

[pbhandar2]

Closing this discussion as it has been inactive. If this is still relevant, then please create a new discussion post and we will get back to you as soon as we can. Thank you!