3 ms·

Very strange to see few to no references to io_uring here. I guess it's still too new. As I've seen many times before so much complexity is replicated in usersp

Very strange to see few to no references to io_uring here. I guess it's still too new. As I've seen many times before so much complexity is replicated in userspace to reproduce kernel behavior out of mmap or DIO/AIO, in order to break the latency, caching, and prioritization into a micromanaged state tuned for a narrow set of applications... Then applied to database code used in a myriad of applications which violate those assumptions and have their own needs. io_uring can't take over fast enough.

jandrewrogers 6y ago
Your assumption is correct, io_uring is too new, it isn't available in most LTS kernels. Give it a few years. Also, if you already have a competent io_submit/O_DIRECT implementation then there are few material performance benefits to io_uring for databases. It mostly just cleans up the API. This has value from a code design/maintenance standpoint, particularly since io_submit is lacking in the documentation department, but the lack of kernel support in most environments makes it a poor tradeoff at this time.
- jlokier 6y ago
  Is it not the case that io_submit can block in some filesystems to handle filesystem metadata (such as block allocation), even with O_DIRECT, whereas io_uring never blocks the submitting thread?
  - jandrewrogers 6y ago
    Yes, in theory. In practice, the way io_submit() is actually used in most systems today would not have that issue, and it is designed that way for other practical reasons. You'd want to use io_uring in a similar way. Even if you ignore the blocking aspect, file system metadata modification at runtime is an edge case factory. For database-y type software generally, it is increasingly uncommon to even install a file system. You work with the raw block devices directly, virtualized or otherwise.

3 ms·

Very strange to see few to no references to io_uring here. I guess it's still too new. As I've seen many times before so much complexity is replicated in usersp

by boxfire 6y ago

jandrewrogers 6y ago
Your assumption is correct, io_uring is too new, it isn't available in most LTS kernels. Give it a few years. Also, if you already have a competent io_submit/O_DIRECT implementation then there are few material performance benefits to io_uring for databases. It mostly just cleans up the API. This has value from a code design/maintenance standpoint, particularly since io_submit is lacking in the documentation department, but the lack of kernel support in most environments makes it a poor tradeoff at this time.
- jlokier 6y ago
  Is it not the case that io_submit can block in some filesystems to handle filesystem metadata (such as block allocation), even with O_DIRECT, whereas io_uring never blocks the submitting thread?
  - jandrewrogers 6y ago
    Yes, in theory. In practice, the way io_submit() is actually used in most systems today would not have that issue, and it is designed that way for other practical reasons. You'd want to use io_uring in a similar way. Even if you ignore the blocking aspect, file system metadata modification at runtime is an edge case factory. For database-y type software generally, it is increasingly uncommon to even install a file system. You work with the raw block devices directly, virtualized or otherwise.