Merge pull request #70 from eigerco/feat/63/benchmarks-for-the-mater-…

…library

feat/36/add benchmarks

Cargo.toml

@@ -40,6 +40,7 @@ cid = { version = "0.11.1" }
 clap = { version = "4.5.3" }
 codec = { package = "parity-scale-codec", version = "3.0.0", default-features = false }
 color-print = "0.3.4"
+criterion = "0.5.1"
 digest = "0.10.7"
 futures = "0.3.28"
 hex-literal = { version = "0.4.1" }

storage/mater/BENCHMARK.md

@@ -0,0 +1,40 @@
+### How to run
+
+Execute the benchmarks with `cargo bench`.
+
+### Results
+
+The benchmarks below use Median times after 100 runs. The duplication
+percentages show the proportion of duplicated content in the file. The
+benchmarks were performed on a machine with a `Ryzen 9 5950X` processor and
+`64GB DDR4` memory.
+
+#### read
+
+Benchmark checks what is the time needed to fully read a content buffer into the `BlockStore`.
+
+| Size / Duplication | 0%        | 10%       | 20%       | 40%       | 80%       |
+| ------------------ | --------- | --------- | --------- | --------- | --------- |
+| 10 MB              | 4.6776 ms | 4.5806 ms | 4.6977 ms | 4.5534 ms | 4.5038 ms |
+| 100 MB             | 62.419 ms | 60.895 ms | 59.461 ms | 55.355 ms | 46.792 ms |
+| 1 GB               | 632.34 ms | 650.01 ms | 631.49 ms | 600.01 ms | 505.58 ms |
+
+#### write
+
+Checks the time needed to write the CARv2 to the buffer from `BlockStore`.
+
+| Size / Duplication | 0%        | 10%       | 20%       | 40%       | 80%       |
+| ------------------ | --------- | --------- | --------- | --------- | --------- |
+| 10 MB              | 1.6516 ms | 1.0342 ms | 875.68 µs | 772.26 µs | 354.77 µs |
+| 100 MB             | 12.689 ms | 10.707 ms | 9.4533 ms | 6.7805 ms | 1.7487 ms |
+| 1 GB               | 123.34 ms | 102.39 ms | 91.712 ms | 69.273 ms | 23.140 ms |
+
+#### filestore
+
+Converts a source file to the CARv2 and writes it to the output file.
+
+| Size / Duplication | 0%        | 10%       | 20%       | 40%       | 80%       |
+| ------------------ | --------- | --------- | --------- | --------- | --------- |
+| 10 MB              | 15.145 ms | 15.179 ms | 15.162 ms | 14.501 ms | 14.836 ms |
+| 100 MB             | 203.85 ms | 210.14 ms | 220.38 ms | 216.34 ms | 211.12 ms |
+| 1 GB               | 1.7674 s  | 1.8174 s  | 1.8396 s  | 1.8496 s  | 1.8774 s  |

storage/mater/Cargo.toml

@@ -23,13 +23,18 @@ serde = { workspace = true, features = ["derive"] }
 serde_ipld_dagcbor.workspace = true
 sha2.workspace = true
 thiserror.workspace = true
-tokio = { workspace = true, features = ["fs", "macros", "rt"] }
+tokio = { workspace = true, features = ["fs", "macros", "rt-multi-thread"] }
 tokio-stream.workspace = true
 tokio-util = { workspace = true, features = ["io"] }
 
 [dev-dependencies]
+criterion = { workspace = true, features = ["async_tokio", "html_reports"] }
 rand.workspace = true
 tempfile.workspace = true
 
 [lints]
 workspace = true
+
+[[bench]]
+harness = false
+name = "benchmark"

storage/mater/README.md

@@ -4,9 +4,9 @@ A Rust library to read and write CAR files.
 
 ## Specifications
 
-CARv1 specification: https://ipld.io/specs/transport/car/carv1/
-CARv2 specification: https://ipld.io/specs/transport/car/carv2/
-UnixFS specification: https://github.com/ipfs/specs/blob/e4e5754ad4a4bfbb2ebe63f4c27631f573703de0/UNIXFS.md
+- CARv1 specification: https://ipld.io/specs/transport/car/carv1/
+- CARv2 specification: https://ipld.io/specs/transport/car/carv2/
+- UnixFS specification: https://github.com/ipfs/specs/blob/e4e5754ad4a4bfbb2ebe63f4c27631f573703de0/UNIXFS.md
 
 ## Developing
 
@@ -26,19 +26,23 @@ The file was generated and checked-in instead of making `pb-rs` part of the buil
 because the definition file ([`unixfs.proto`](src/unixfs/unixfs.proto)) does not
 change frequently, hence, there is no need to add complexity to the build process.
 
+### Benchmarks
+
+[Read more](BENCHMARK.md)
+
 ## Acknowledgements
 
 We'd like to thank all the people that participated in the projects mentioned in this section.
 In a way or another, they were all instrumental in the implementation of the present library.
 
-* [go-car](https://github.com/ipld/go-car) — the original implementation.
-* [beetle](https://github.com/n0-computer/beetle) — the library `mater` is based on.
+- [go-car](https://github.com/ipld/go-car) — the original implementation.
+- [beetle](https://github.com/n0-computer/beetle) — the library `mater` is based on.
   We've gutted out the important bits for this project, but without it, this work would've been much harder.
-* [ImHex](https://github.com/WerWolv/ImHex) — for saving hours when comparing binary files.
+- [ImHex](https://github.com/WerWolv/ImHex) — for saving hours when comparing binary files.
 
 ### Similar libraries/sources
 
-* [Forest](https://github.com/ChainSafe/forest/blob/62e55df27a091ba7993a60cc1e72622ad8e25151/src/utils/db/car_stream.rs#L155)
-* [rust-car](https://github.com/jaeaster/rust-car)
-* [rs-car](https://github.com/dapplion/rs-car)
-* [car-utils](https://github.com/blocklessnetwork/car-utils)
+- [Forest](https://github.com/ChainSafe/forest/blob/62e55df27a091ba7993a60cc1e72622ad8e25151/src/utils/db/car_stream.rs#L155)
+- [rust-car](https://github.com/jaeaster/rust-car)
+- [rs-car](https://github.com/dapplion/rs-car)
+- [car-utils](https://github.com/blocklessnetwork/car-utils)