stencillogic / db-core Goto Github PK

Trigger buffer writers inside BlockMgr::allocate_on_cache when not enough space in buffer left. Wait untill dirty blocks are written
and retry allocation. Repeat until successful.

Implement unit tests for module crate::tran_mgr::tran_mgr

Add unit tests for module crate::storage::block_driver

Add unit tests for module crate::system::checkpointer

Add unit tests for modules crate::log_mgr

Currently deleted records are not vacuumed. Vaccuming must occur to free up space in data blocks as soon as deleted entries are not needed anymore.
Implementation: TBD

Currently, objects are collected withing files by specifying file_id when creating an object. To add support of block device file must be replaced with some other kind of object grouping.
Possible implementation: link extents of a single object grouping in the linked list instead of grouping extents by files.

More tests should be added to test Instance under different conditions. E.g. intensive concurrent reads and writes, transactions competing for resources, writing large objects or large number of small objects, which involves allocation of new extents, etc.