PersistedSmtForest

[gswirski]

Draft PR to explore persistent SmtForest version.

LargeSmt baseline

Running benchmark with rocksdb storage
Benchmarking LargeSmt implementation
Running a construction benchmark:
Using database path: /var/folders/h4/sk46p_vd4_7fwnwr9989wz4c0000gn/T/miden_crypto_benchmark
Constructed an SMT with 100000 key-value pairs in 2.1 seconds
Number of leaf nodes: 100000

Running an insertion benchmark:
The average insertion time measured by 10000 inserts into an SMT with 100000 leaves is 355 μs

Running a batched insertion benchmark:
The average insert-batch computation time measured by a 10000-batch into an SMT with 110000 leaves over 203.3 ms is 20 μs
The average insert-batch application time measured by a 10000-batch into an SMT with 110000 leaves over 147.2 ms is 15 μs
The average batch insertion time measured by a 10000-batch into an SMT with 110000 leaves totals to 350.5 ms

Running a batched update benchmark:
The average update-batch computation time measured by a 10000-batch into an SMT with 120000 leaves over 205.7 ms is 21 μs
The average update-batch application time measured by a 10000-batch into an SMT with 120000 leaves over 144.6 ms is 14 μs
The average batch update time measured by a 10000-batch into an SMT with 120000 leaves totals to 350.3 ms

Running a proof generation benchmark:
The average proving time measured by 100 value proofs in an SMT with 117982 leaves in 8 μs
Benchmark completed successfully

Naive implementation

Each node hash stored as a key in rocksdb.

• creation: 17x slower
• batch insert: 13x slower
• batch update: 13x slower
• proof generation: 20x slower

Running `target/release/miden-crypto --storage rocksdb --size 100000 --tree persisted-forest`
Running benchmark with rocksdb storage
Benchmarking PersistedSmtForest implementation
Running a construction benchmark:
Using forest database path: /var/folders/h4/sk46p_vd4_7fwnwr9989wz4c0000gn/T/miden_crypto_persisted_forest_benchmark
Constructed an SMT with 100000 key-value pairs in 35.0 seconds
Number of leaf nodes: 100000

Running an insertion benchmark:
The average insertion time measured by 10000 inserts into an SMT with 100000 leaves is 537 μs

Running a batched insertion benchmark:
The average insert-batch execution time measured by a 10000-batch into an SMT with 110000 leaves over 4638.4 ms is 464 μs
The average batch insertion time measured by a 10000-batch into an SMT with 110000 leaves totals to 4638.4 ms

Running a batched update benchmark:
The average update-batch execution time measured by a 10000-batch into an SMT with 120000 leaves over 4454.8 ms is 445 μs
The average batch update time measured by a 10000-batch into an SMT with 120000 leaves totals to 4454.8 ms

Running a proof generation benchmark:
The average proving time measured by 100 value proofs in an SMT with 117996 leaves in 201 μs
Benchmark completed successfully