sosp23-s3fifo
golang-fifo
sosp23-s3fifo | golang-fifo | |
---|---|---|
2 | 4 | |
83 | 114 | |
- | - | |
6.6 | 8.6 | |
8 months ago | about 2 months ago | |
C | Go | |
- | MIT License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
sosp23-s3fifo
-
Otter, Fastest Go in-memory cache based on S3-FIFO algorithm
We observed that quick demotion[2] is important to achieve a low miss ratio in modern cache workloads, and existing algorithms such as TinyLFU and LIRS have lower miss ratios because of the small 1% window they use. This motivated us to design S3-FIFO, which uses simple FIFO queues to achieve low miss ratios. It is true that compared to state-of-the-art, S3-FIFO does not use any fancy techniques, but this does not mean it has bad performance.
In our large-scale evaluations, we found that the fancy techniques in LIRS, ARC, and TinyLFU can sometimes increase the miss ratio. But simple FIFO queues are more robust. However, *it is not true that S3-FIFO is better on every trace*.
* Note that some of the S3-FIFO results in Otter's repo are not updated and have an implementation bug, and we are working with the owner to update them.
[1] https://github.com/Thesys-lab/sosp23-s3fifo?tab=readme-ov-fi...
golang-fifo
-
Otter, Fastest Go in-memory cache based on S3-FIFO algorithm
Hello, Thank you for replying here :)
Many of answers you replied are reasonable and good.
And I just want to add more comments for others.
1. SIEVE is not scan-resistant, so that, I think it should only be applied for web cache workloads (typlically follows power-law distribution)
2. SIEVE is somewhat scalalbe for read-intensive applications (e.g. blog, shop and etc), because it doesn't require to hold a lock on cahce hit.
3. The purpose of golang-fifo is to provide simple and efficient cache implementation (e.g. hashicorp-lru, groupcache)
4. when increasing contention otter sacrifices 1-2 percent
-> I think that the statement is incorrect. The hit rate varies depending on the total number of objects and the size of the cache, so it should be compared relatively. for example, otter's efficiency decreased by 5% compared to single-threaded when lock contention increased (decreased efficiency makes a mean network latency higher, because it may need to conduct heavy operation e.g. re-calculation, database access and so on)
5. ghost queue : honetly at that time of writing the code, I didn't deep dive into the bucket table implementation, it may not work same as actual bucket hash table (see here: https://github.com/scalalang2/golang-fifo/issues/16)
-
golang-fifo | Modern cache eviction algorithm implementations.
I'm also implementing cache algorithms, introduced in papers, in golang.you can visit here. Your contribution would be greatly appreciated.
What are some alternatives?
libCacheSim - a high performance library for building cache simulators
otter - A high performance lockless cache for Go.
xsync - Concurrent data structures for Go
theine-go - high performance in-memory cache
ristretto - A high performance memory-bound Go cache