wanghenshui / wanghenshui.github.io

my blog, please do not fork
https://wanghenshui.github.io
Other
4 stars 1 forks source link

filter #109

Open wanghenshui opened 6 months ago

wanghenshui commented 6 months ago

https://github.sheincorp.cn/FastFilter/xor_singleheader

xor binaryfuse,都是静态的。

给rocksdb提个MR?

https://www.modb.pro/db/32203

总结不错,考虑到删除,还得是cuckoo filter

https://bugwz.com/2019/08/12/bloom-and-cuckoo-filter

https://juejin.cn/post/6844903861749055502

删除存在误报问题

https://www.jos.org.cn/html/2023/3/6782.htm

wanghenshui commented 5 months ago

https://www.vldb.org/pvldb/vol13/p197-wang.pdf Vacuum Filters 我服了这哥们起名字能不能考虑重名的问题?

wanghenshui commented 5 months ago

https://yangtonghome.github.io/uploads/Additive_and_Subtractive_Cuckoo_Filters.pdf

wanghenshui commented 5 months ago

https://blog.fleek.network/post/bloom-and-cuckoo-filters-for-cache-summarization/

Conclusion

We examined whether Counting Bloom filters or Cuckoo filters are more suitable for summarizing caches on Fleek Network. Cuckoo filters are more space-efficient, especially for lower false positive rates. Bloom filters have a slightly better insertion and lookup performance for the implementations we tested.

Both filters can be adapted to grow and shrink in size dynamically. Since the difference in insertion and lookup performance is negligible while Cuckoo filters are significantly more space-efficient, we favor Cuckoo filters for our use case.

没测 binary fuse filter

测试场景可以参考一下