Revert "Add support for caching bloomfilters (#1204)" #1256
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This reverts commit 4676ca9. Reading bloom filters from the cache leads to severe preformance degradation. Result of read bench tool 1. With bloom filters in cache Average read speed : 97.37 KB/s Total bytes read in 1 minute : 3.7 MB 2. With bloom filters stored in memory (no cache) Average read speed : 41.3833 MB/s Total bytes read in 1 minute : 2.4 GB
jarifibrahim
requested review from
ashish-goswami and
manishrjain
as code owners
This was referenced Mar 12, 2020
jarifibrahim
pushed a commit
that referenced
this pull request
Mar 16, 2020
#1256 showed that a single cache might not be enough to store the data blocks and the bloom filters. This commit adds a separate cache for the bloom filters. This commit also adds a new flag `LoadBloomsOnOpen` which determines if the bloom filters should be loaded when the table is opened on or not. The default value of `MaxBfCacheSize` is `zero` and `LoadBloomsOnOpen` is true. This change has significant performance improvement on read speeds because a single cache would lead to bloom filter eviction and we would read the bloom filter from the disk.
|
This PR is no longer necessary. The issue has been fixed by #1260 |
jarifibrahim
pushed a commit
that referenced
this pull request
Mar 24, 2020
#1256 showed that a single cache might not be enough to store the data blocks and the bloom filters. This commit adds a separate cache for the bloom filters. This commit also adds a new flag `LoadBloomsOnOpen` which determines if the bloom filters should be loaded when the table is opened on or not. The default value of `MaxBfCacheSize` is `zero` and `LoadBloomsOnOpen` is true. This change has significant performance improvement on read speeds because a single cache would lead to bloom filter eviction and we would read the bloom filter from the disk. (cherry picked from commit eaf64c0)
manishrjain
pushed a commit
to outcaste-io/outserv
that referenced
this pull request
Jul 6, 2022
hypermodeinc/badger#1256 showed that a single cache might not be enough to store the data blocks and the bloom filters. This commit adds a separate cache for the bloom filters. This commit also adds a new flag `LoadBloomsOnOpen` which determines if the bloom filters should be loaded when the table is opened on or not. The default value of `MaxBfCacheSize` is `zero` and `LoadBloomsOnOpen` is true. This change has significant performance improvement on read speeds because a single cache would lead to bloom filter eviction and we would read the bloom filter from the disk.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This reverts commit 4676ca9.
Reading bloom filters from the cache leads to severe performance degradation.
Result of read bench tool with
key size 32 bytesandvalue size 128 bytes(Note: data is not compressed or encrypted)The performance degradation is a result of the way ristretto handles item eviction from the cache. For instance, the following was the state of cache after opening 65 tables
The important metric here is the
keys-added: 194 and keys-evicted: 172. This means we have only22items in the cache, which includes the bloom filters and the sst blocks. So most of the time the bloom filter has to be read from the SST which is very expensive.Fixes #1255, #1254 and #1248
This change is