Struct BlockEngineConfig 

pub struct BlockEngineConfig<K, V, P>
where
    K: StorageKey,
    V: StorageValue,
    P: Properties,
{ /* private fields */ }

Config for the block-based disk cache engine.

The block-based disk cache engine is suitable for general cache entries with size from 2K to hundreds of MiBs.

Each cache entry will be aligned to a multiplier of 4K on disk, hence too small cache entries will lead to heavy internal fragmentation.

The disk cache evicts cache entries in block unit.

Implementations

impl<K, V, P> BlockEngineConfig<K, V, P>

where K: StorageKey, V: StorageValue, P: Properties,

pub fn new(device: Arc<dyn Device>) -> Self

Create a new block-based disk cache engine builder with default configurations.

pub fn with_block_size(self, block_size: usize) -> Self

Set the block size for the block-based disk cache engine.

Block is the minimal cache eviction unit for the block-based disk cache, its size also limits the max cacheable entry size.

The block size must be 4K-aligned. the given value is not 4K-aligned, it will be automatically aligned up.

Default: 16 MiB.

pub fn with_indexer_shards(self, indexer_shards: usize) -> Self

Set the shard num of the indexer. Each shard has its own lock.

Default: 64.

pub fn with_recover_concurrency(self, recover_concurrency: usize) -> Self

Set the recover concurrency for the disk cache store.

Default: 8.

pub fn with_flushers(self, flushers: usize) -> Self

Set the flusher count for the disk cache store.

The flusher count limits how many blocks can be concurrently written.

Default: 1.

pub fn with_admission_filter(self, filter: StorageFilter) -> Self

Set the admission filter for th disk cache store.

The admission filter is used to pick the entries that can be inserted into the disk cache store.

Default: Admit all.

pub fn with_reclaimers(self, reclaimers: usize) -> Self

Set the reclaimer count for the disk cache store.

The reclaimer count limits how many blocks can be concurrently reclaimed.

Default: 1.

pub fn with_buffer_pool_size(self, buffer_pool_size: usize) -> Self

Set the total flush buffer pool size.

Each flusher shares a volume at threshold / flushers.

If the buffer of the flush queue exceeds the threshold, the further entries will be ignored.

Default: 16 MiB.

pub fn with_blob_index_size(self, blob_index_size: usize) -> Self

Set the blob index size for each blob.

A larger blob index size can hold more blob entries, but it will also increase the io size of each blob part write.

NOTE: The size will be aligned up to a multiplier of 4K.

Default: 4 KiB

pub fn with_submit_queue_size_threshold( self, submit_queue_size_threshold: usize, ) -> Self

Set the submit queue size threshold.

If the total entry estimated size in the submit queue exceeds the threshold, the further entries will be ignored.

Default: buffer_pool_size * 2.

pub fn with_clean_block_threshold(self, clean_block_threshold: usize) -> Self

Set the clean block threshold for the disk cache store.

The reclaimers only work when the clean block count is equal to or lower than the clean block threshold.

Default: the same value as the reclaimers.

pub fn with_eviction_pickers( self, eviction_pickers: Vec<Box<dyn EvictionPicker>>, ) -> Self

Set the eviction pickers for th disk cache store.

The eviction picker is used to pick the block to reclaim.

The eviction pickers are applied in order. If the previous eviction picker doesn’t pick any block, the next one will be applied.

If no eviction picker picks a block, a block will be picked randomly.

Default: [ invalid ratio picker { threshold = 0.8 }, fifo picker ]

pub fn with_reinsertion_filter(self, filter: StorageFilter) -> Self

Set the reinsertion filter for th disk cache store.

The reinsertion filter is used to pick the entries that can be reinsertion into the disk cache store while reclaiming.

Note: Only extremely important entries should be picked. If too many entries are picked, both insertion and reinsertion will be stuck.

Default: Reject all.

pub fn with_tombstone_log(self, enable: bool) -> Self

Enable the tombstone log.

For updatable cache, either the tombstone log or crate::engine::RecoverMode::None must be enabled to prevent from the phantom entries after reopen.

pub async fn build( self: Box<Self>, __arg1: EngineBuildContext, ) -> Result<Arc<BlockEngine<K, V, P>>>

Build the block-based disk cache engine with the given configurations.

Trait Implementations

impl<K, V, P> Debug for BlockEngineConfig<K, V, P>

where K: StorageKey, V: StorageValue, P: Properties,

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<K, V, P> EngineConfig<K, V, P> for BlockEngineConfig<K, V, P>

where K: StorageKey, V: StorageValue, P: Properties,

fn build( self: Box<Self>, ctx: EngineBuildContext, ) -> BoxFuture<'static, Result<Arc<dyn Engine<K, V, P>>>>

Build the engine with the given configurations.

fn boxed(self) -> Box<Self>

where Self: Sized,

Box the config.

impl<K, V, P> From<BlockEngineConfig<K, V, P>> for Box<dyn EngineConfig<K, V, P>>

where K: StorageKey, V: StorageValue, P: Properties,

fn from(builder: BlockEngineConfig<K, V, P>) -> Self

Converts to this type from the input type.