Caching libraries Benchmarks

According to my new library cachebox, I decided to benchmark caching libraries which are I know, to show the power of cachebox ...

If you know other library, tell me to add it to this page.

Note

The cacheing.VTTLCache class was excluded from the benchmark because assigning a key raises TypeError.

Methodology

Times are from Python: each sample is a calibrated inner loop around a single API call (set / get / delete / popitem, or one entry of an items() iteration). Setup and teardown are not timed. GC is collected before each sample and disabled during the timed region.

Reported values are the median ns/op (tables ranked by median), plus mean, stdev, MAD, ops/sec, sample count, and coefficient of variation. Outliers are not dropped. A result is flagged when CV > 10% or max > 1.5× mean.

All implementations use the same working set (maxsize = 4096, integer keys/values). TTL caches are constructed with a 3600s TTL so nothing expires during a sample. dict, cachebox.Cache, and cachetools.Cache have no eviction policy and are skipped on eviction cases.

Case	What is measured
Set new (no eviction)	Insert keys that are not present and do not evict (cache ~25% full).
Set at capacity (eviction)	Insert new keys into a cache already at maxsize.
Lookup existing	`get` of a present key (LRU/LFU recency/frequency updates included).
Lookup missing	`get` of a missing key.
Deletion	Delete present keys (one delete per populated key, then a new cache).
Popitem	`popitem()` on a cache filled to maxsize.
Hash collision	`get` of keys whose `hash()` is always 42.
Iterate items	`for k, v in cache.items()` over a cache filled to maxsize; time is per entry.

moka-py is a concurrent Rust cache measured single-threaded like the others.

🖥️ System Information

Platform: Linux-7.2.11-arch1-1-x86_64-with-glibc2.44
Python: 3.14.8
Processor: 11th Gen Intel(R) Core(TM) i3-1115G4 @ 3.00GHz
Timer: resolution 263.0 ns, overhead 42.0 ns
Run: min-time 100 ms/sample, warmup 1, samples 25, isolated=no, fast=no, pin-cpu=off, nice=False
Hash seed: unset (isolated workers use 0 except the collision case)
GC: disabled during the timed loop; collected before each sample
Threading: single-threaded (including moka-py)
Libraries:
- cachebox: 6.2.9
- cachetools: 7.2.7
- cacheing: 0.1.1
- lru-dict: 1.4.1
- moka-py: 0.5.0
- matplotlib: 3.11.2
- matplotx: 0.3.10

📊 Set new (no eviction)

Insert keys that do not exist and do not trigger eviction. Cache is pre-filled to 25% of maxsize.

Range: 16.7× between fastest and slowest (median)
Best: dict (66 ns, 14.9M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
dict	66.0	67.2	3.7	0.5	14.9M	25	5.6%
cachebox.Cache	99.7	100.7	5.1	1.9	9.9M	200	5.0%
cachebox.RRCache	108.7	109.2	3.7	2.0	9.2M	200	3.3%
cachebox.FIFOCache	112.2	113.1	5.2	1.7	8.8M	200	4.6%
cachebox.LRUCache	134.6	136.0	7.6	2.0	7.4M	200	5.6%
cachebox.LFUCache	136.3	136.8	5.8	1.8	7.3M	200	4.2%
lru.LRU	161.0	162.1	7.5	2.1	6.2M	200	4.6%
cachebox.TTLCache	162.0	163.0	5.6	2.2	6.1M	200	3.5%
cachebox.VTTLCache	168.4	170.7	10.1	2.3	5.9M	200	5.9%
cachetools.Cache	198.2	199.1	6.6	2.9	5.0M	200	3.3%
cacheing.RandomCache	297.3	300.7	12.1	3.5	3.3M	200	4.0%
cachetools.LFUCache	307.3	308.6	7.8	5.0	3.2M	200	2.5%
cachetools.FIFOCache	308.2	309.8	8.7	4.0	3.2M	200	2.8%
cachetools.RRCache	315.1	320.8	33.3	5.0	3.1M	200	10.4%
cacheing.LRUCache	334.0	336.7	11.6	3.6	3.0M	200	3.4%
cacheing.LFUCache	337.4	340.9	13.4	4.9	2.9M	200	3.9%
moka_py.Moka	402.0	403.2	10.4	4.9	2.5M	200	2.6%
moka_py.Moka (lru)	416.0	417.4	11.3	6.1	2.4M	200	2.7%
cachetools.LRUCache	510.5	512.9	10.1	5.7	1.9M	200	2.0%
cachetools.TTLCache	756.4	783.4	99.4	6.5	1.3M	200	12.7%
cacheing.TTLCache	1,102.9	1,122.8	68.6	12.7	0.9M	200	6.1%

📊 Set at capacity (eviction)

Insert new keys into a cache already at maxsize so each set evicts an entry.

Range: 16.6× between fastest and slowest (median)
Best: cachebox.LRUCache (110 ns, 8.9M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
cachebox.LRUCache	109.8	112.3	13.3	3.0	8.9M	200	11.8%
cachebox.FIFOCache	118.0	118.5	5.1	1.1	8.4M	200	4.3%
cachebox.RRCache	139.1	142.8	20.4	2.4	7.0M	200	14.3%
cachebox.TTLCache	173.0	173.8	5.8	2.4	5.8M	200	3.3%
lru.LRU	189.6	193.2	23.2	2.5	5.2M	200	12.0%
cachebox.LFUCache	203.3	206.6	21.2	3.8	4.8M	200	10.3%
cachebox.VTTLCache	247.9	251.6	25.3	4.8	4.0M	200	10.0%
cacheing.LRUCache	597.6	607.4	24.7	5.3	1.6M	200	4.1%
moka_py.Moka	676.4	695.5	66.2	7.1	1.4M	163	9.5%
cachetools.FIFOCache	713.8	750.9	121.7	25.5	1.3M	200	16.2%
cacheing.LFUCache	736.2	741.2	15.4	3.4	1.3M	108	2.1%
moka_py.Moka (lru)	758.4	782.1	70.4	7.8	1.3M	191	9.0%
cachetools.RRCache	992.9	1,015.7	96.0	8.1	1.0M	89	9.5%
cachetools.LRUCache	999.6	1,024.2	93.7	13.3	1.0M	151	9.1%
cacheing.RandomCache	1,017.1	1,083.7	173.1	10.9	0.9M	143	16.0%
cacheing.TTLCache	1,511.1	1,554.2	156.9	11.4	0.6M	125	10.1%
cachetools.LFUCache	1,771.5	1,807.1	117.0	6.2	0.6M	194	6.5%
cachetools.TTLCache	1,821.3	1,849.2	121.6	20.8	0.5M	107	6.6%

Skipped: dict (no eviction policy), cachebox.Cache (no eviction policy), cachetools.Cache (no eviction policy)

📊 Lookup existing

Look up keys that are in the cache. LRU/LFU recency or frequency updates are part of the API cost.

Range: 13.6× between fastest and slowest (median)
Best: dict (79 ns, 12.2M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
dict	78.7	82.1	9.9	0.5	12.2M	25	12.1%
cachebox.Cache	93.2	93.2	0.6	0.3	10.7M	25	0.7%
cachebox.RRCache	93.5	93.5	0.6	0.3	10.7M	25	0.6%
cachebox.FIFOCache	95.7	95.5	0.6	0.3	10.5M	25	0.6%
cachebox.LFUCache	96.5	96.6	0.7	0.6	10.3M	25	0.7%
cachebox.LRUCache	99.7	99.8	0.5	0.5	10.0M	25	0.5%
cacheing.RandomCache	109.6	109.9	1.5	0.4	9.1M	25	1.4%
cachebox.VTTLCache	127.3	127.4	0.6	0.5	7.9M	25	0.4%
cachebox.TTLCache	128.5	128.5	0.5	0.5	7.8M	25	0.4%
lru.LRU	135.4	138.4	15.4	0.1	7.2M	25	11.1%
cachetools.Cache	171.7	171.8	0.4	0.2	5.8M	25	0.2%
cachetools.RRCache	173.1	173.1	0.4	0.4	5.8M	25	0.3%
cachetools.FIFOCache	173.2	173.3	0.7	0.3	5.8M	25	0.4%
cacheing.LRUCache	201.0	201.2	1.3	0.6	5.0M	25	0.6%
moka_py.Moka (lru)	237.3	237.7	1.1	0.3	4.2M	25	0.4%
moka_py.Moka	246.0	246.1	0.9	0.5	4.1M	25	0.4%
cachetools.LRUCache	262.7	262.7	0.5	0.3	3.8M	25	0.2%
cacheing.LFUCache	372.3	370.7	5.8	2.7	2.7M	25	1.6%
cachetools.LFUCache	417.1	420.4	11.0	1.1	2.4M	25	2.6%
cacheing.TTLCache	786.2	787.6	6.3	1.6	1.3M	25	0.8%
cachetools.TTLCache	1,068.9	1,078.2	45.3	4.5	0.9M	25	4.2%

📊 Lookup missing

Look up keys that are not in a populated cache (mapping.get with default None).

Range: 20.0× between fastest and slowest (median)
Best: dict (83 ns, 12.1M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
dict	82.7	82.9	0.6	0.3	12.1M	25	0.7%
cachebox.LRUCache	92.4	92.7	1.2	0.7	10.8M	25	1.3%
cachebox.Cache	93.5	93.5	0.5	0.4	10.7M	25	0.5%
cachebox.LFUCache	93.9	93.8	0.9	0.8	10.7M	25	1.0%
cachebox.RRCache	94.4	94.6	0.8	0.5	10.6M	25	0.8%
cachebox.FIFOCache	94.5	94.3	0.8	0.6	10.6M	25	0.8%
cachebox.VTTLCache	104.4	105.1	2.4	0.6	9.5M	25	2.3%
cachebox.TTLCache	109.7	109.7	0.4	0.3	9.1M	25	0.4%
cachetools.LRUCache	133.5	133.8	1.0	0.4	7.5M	25	0.8%
cachetools.RRCache	133.7	133.9	0.8	0.2	7.5M	25	0.6%
cachetools.Cache	133.9	133.8	0.6	0.5	7.5M	25	0.4%
cachetools.LFUCache	134.3	134.2	0.6	0.3	7.5M	25	0.4%
cachetools.FIFOCache	134.7	134.6	0.9	0.9	7.4M	25	0.7%
moka_py.Moka (lru)	170.6	170.6	0.5	0.4	5.9M	25	0.3%
moka_py.Moka	179.7	179.7	0.4	0.3	5.6M	25	0.2%
lru.LRU	195.8	196.0	1.0	0.4	5.1M	25	0.5%
cacheing.RandomCache	480.0	480.2	5.6	4.2	2.1M	25	1.2%
cacheing.LFUCache	571.6	572.4	4.6	1.4	1.7M	25	0.8%
cachetools.TTLCache	655.5	666.7	22.3	3.1	1.5M	25	3.3%
cacheing.LRUCache	703.7	706.2	8.2	2.0	1.4M	25	1.2%
cacheing.TTLCache	1,652.5	1,645.5	25.7	17.9	0.6M	25	1.6%

📊 Deletion

Delete keys that exist. Each sample populates maxsize entries and deletes them once.

Range: 10.0× between fastest and slowest (median)
Best: dict (71 ns, 14.0M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
dict	71.1	71.3	1.8	1.1	14.0M	200	2.5%
cachebox.Cache	93.7	94.1	2.6	1.5	10.6M	200	2.8%
cachebox.FIFOCache	94.5	94.9	3.8	1.4	10.5M	200	4.0%
cachebox.RRCache	96.8	97.4	6.5	1.6	10.3M	200	6.7%
cachebox.LRUCache	106.1	106.8	5.1	2.3	9.4M	200	4.8%
lru.LRU	108.2	109.4	7.1	2.2	9.1M	200	6.4%
cachebox.TTLCache	124.8	127.0	12.0	2.3	7.9M	200	9.4%
cachetools.Cache	146.5	146.8	3.1	2.0	6.8M	200	2.1%
cachebox.LFUCache	180.3	182.7	18.4	2.8	5.5M	200	10.0%
cachebox.VTTLCache	210.2	212.2	11.8	4.1	4.7M	200	5.6%
cacheing.LRUCache	230.2	243.6	47.9	4.8	4.1M	200	19.6%
cachetools.FIFOCache	239.2	253.0	38.3	6.0	4.0M	200	15.1%
cachetools.LRUCache	240.9	257.1	39.5	6.9	3.9M	200	15.4%
cachetools.LFUCache	273.6	281.2	34.6	6.3	3.6M	200	12.3%
cachetools.RRCache	334.9	342.1	26.8	7.6	2.9M	200	7.8%
cacheing.LFUCache	339.7	370.4	75.8	26.4	2.7M	200	20.5%
cacheing.RandomCache	368.8	397.0	71.2	23.1	2.5M	200	17.9%
cachetools.TTLCache	631.6	603.8	44.8	19.1	1.7M	83	7.4%
moka_py.Moka	682.3	688.9	21.9	5.2	1.5M	144	3.2%
moka_py.Moka (lru)	684.9	686.1	12.2	4.1	1.5M	127	1.8%
cacheing.TTLCache	710.4	721.6	63.1	7.7	1.4M	200	8.7%

📊 Popitem

Call popitem() on a cache filled to maxsize (the eviction primitive).

Range: 13.1× between fastest and slowest (median)
Best: cachebox.FIFOCache (102 ns, 9.8M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
cachebox.FIFOCache	102.0	101.7	3.1	2.1	9.8M	200	3.0%
cachebox.TTLCache	108.1	109.1	5.2	2.0	9.2M	200	4.8%
cachebox.LRUCache	110.9	112.2	7.3	2.4	8.9M	200	6.5%
lru.LRU	137.0	137.8	6.9	2.1	7.3M	200	5.0%
cachebox.RRCache	148.9	152.0	12.1	4.9	6.6M	200	8.0%
cachebox.LFUCache	179.9	180.2	6.7	3.3	5.5M	200	3.7%
cachebox.VTTLCache	188.3	190.0	9.1	3.8	5.3M	200	4.8%
cacheing.LRUCache	242.7	246.0	12.2	5.4	4.1M	200	5.0%
cacheing.TTLCache	357.0	364.0	24.0	4.3	2.7M	200	6.6%
cacheing.LFUCache	358.5	364.1	16.6	7.1	2.7M	200	4.6%
cachetools.FIFOCache	447.0	447.5	50.8	41.3	2.2M	200	11.3%
cachetools.LRUCache	453.6	447.3	43.0	35.1	2.2M	200	9.6%
cachetools.RRCache	645.3	658.3	40.3	7.1	1.5M	145	6.1%
cacheing.RandomCache	657.7	683.1	70.2	11.4	1.5M	181	10.3%
cachetools.TTLCache	1,037.8	1,095.1	169.1	9.5	0.9M	162	15.4%
cachetools.LFUCache	1,333.8	1,344.9	33.4	7.6	0.7M	200	2.5%

Skipped: dict (no eviction policy), cachebox.Cache (no eviction policy), cachetools.Cache (no eviction policy), moka_py.Moka (no popitem), moka_py.Moka (lru) (no popitem)

📊 Hash collision

Look up keys that all hash to the same value (hash() == 42).

Range: 7.0× between fastest and slowest (median)
Best: cachebox.RRCache (1002 ns, 1.0M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
cachebox.RRCache	1,002.3	1,005.8	6.7	2.6	1.0M	25	0.7%
cachebox.Cache	1,005.1	1,010.6	15.3	2.1	1.0M	25	1.5%
cachebox.LRUCache	1,043.5	1,051.4	20.7	3.5	1.0M	25	2.0%
cachebox.LFUCache	1,050.8	1,056.5	8.9	2.9	0.9M	25	0.8%
cachebox.VTTLCache	1,071.5	1,109.4	63.7	1.3	0.9M	25	5.7%
dict	1,182.1	1,188.2	11.4	1.1	0.8M	25	1.0%
cachebox.FIFOCache	1,187.3	1,248.8	106.6	1.9	0.8M	25	8.5%
cacheing.RandomCache	1,206.0	1,220.7	39.7	1.0	0.8M	25	3.3%
cachebox.TTLCache	1,214.9	1,229.7	26.3	2.4	0.8M	25	2.1%
lru.LRU	1,237.1	1,245.8	25.1	0.8	0.8M	25	2.0%
moka_py.Moka	2,126.2	2,190.6	121.8	5.7	0.5M	25	5.6%
moka_py.Moka (lru)	2,134.3	2,220.1	180.0	3.8	0.5M	25	8.1%
cachetools.Cache	2,361.1	2,371.6	28.8	1.0	0.4M	25	1.2%
cachetools.FIFOCache	2,364.0	2,405.1	120.9	1.3	0.4M	25	5.0%
cachetools.RRCache	2,364.3	2,383.1	47.0	2.2	0.4M	25	2.0%
cacheing.LRUCache	2,399.3	2,403.2	11.7	0.7	0.4M	25	0.5%
cacheing.TTLCache	3,077.3	3,086.9	25.7	5.7	0.3M	25	0.8%
cachetools.LRUCache	3,561.1	3,573.3	32.1	1.4	0.3M	25	0.9%
cachetools.TTLCache	5,504.6	5,598.0	257.3	8.6	0.2M	25	4.6%
cacheing.LFUCache	6,522.9	6,575.1	175.6	16.2	0.2M	25	2.7%
cachetools.LFUCache	7,047.8	7,098.0	150.3	2.7	0.1M	25	2.1%

📊 Iterate items

Iterate cache.items() over a cache filled to maxsize. Time is per entry; creating the iterator is included in the timing.

Range: 33.5× between fastest and slowest (median)
Best: cacheing.TTLCache (16 ns, 62.4M ops/sec)

Implementation	Median (ns)	Mean (ns)	Stdev	MAD	Ops/sec	Samples	CV
cacheing.TTLCache	15.9	16.0	0.3	0.1	62.4M	25	1.7%
cacheing.LFUCache	16.1	16.7	1.1	0.1	59.8M	25	6.7%
dict	16.1	16.8	1.1	0.1	59.7M	25	6.7%
cacheing.RandomCache	16.3	16.6	0.7	0.1	60.3M	25	4.3%
cacheing.LRUCache	16.6	17.0	0.9	0.1	58.9M	25	5.5%
lru.LRU	24.7	24.8	0.5	0.1	40.3M	25	1.9%
cachebox.LRUCache	36.7	38.2	2.5	0.2	26.2M	25	6.6%
cachebox.Cache	37.5	38.0	1.4	0.1	26.3M	25	3.6%
cachebox.RRCache	37.5	37.7	0.3	0.0	26.5M	25	0.8%
cachebox.FIFOCache	37.6	37.8	1.7	0.2	26.5M	25	4.5%
cachebox.LFUCache	37.8	37.7	1.7	0.9	26.6M	25	4.4%
cachebox.VTTLCache	50.7	51.9	2.6	0.1	19.3M	25	5.1%
cachebox.TTLCache	50.9	51.5	1.3	0.2	19.4M	25	2.5%
cachetools.Cache	63.8	66.1	5.2	0.3	15.1M	25	7.9%
cachetools.FIFOCache	64.2	64.8	2.1	0.2	15.4M	25	3.3%
cachetools.RRCache	64.8	64.7	0.4	0.1	15.4M	25	0.7%
cachetools.LRUCache	138.5	143.3	12.4	0.3	7.0M	25	8.7%
cachetools.LFUCache	277.9	287.7	24.5	2.7	3.5M	25	8.5%
cachetools.TTLCache	532.3	536.8	12.0	0.6	1.9M	25	2.2%

Skipped: moka_py.Moka (no iteration API), moka_py.Moka (lru) (no iteration API)

🚀 Usage

Prerequisites

uv sync

Run Benchmarks

# All cases (in-process)
python -m cachebox_benchmark

# One or more cases
python -m cachebox_benchmark get_hit delete

# Faster iteration (not for published numbers)
python -m cachebox_benchmark --fast

# One subprocess per (case, implementation)
python -m cachebox_benchmark --isolated

# JSON + regenerate README from a previous run
python -m cachebox_benchmark --json results.json
python -m cachebox_benchmark --report-only results.json

Optional host knobs (off by default): --pin-cpu 0, --nice.

Understanding Results

Median (ns): robust typical cost per operation; tables are sorted by this.
Mean / Stdev / MAD: distribution; MAD is median absolute deviation.
Ops/sec: throughput from the mean.
CV: stdev / mean. High CV means the number is noisy, not that the library is “bad”.

Name		Name	Last commit message	Last commit date
cachebox_benchmark		cachebox_benchmark
photos		photos
.gitignore		.gitignore
.python-version		.python-version
README.md		README.md
pyproject.toml		pyproject.toml
uv.lock		uv.lock

Navigation Menu

Repository navigation

About

Repository files navigation

Caching libraries Benchmarks

Methodology

🖥️ System Information

📊 Set new (no eviction)

📊 Set at capacity (eviction)

📊 Lookup existing

📊 Lookup missing

📊 Deletion

📊 Popitem

📊 Hash collision

📊 Iterate items

🚀 Usage

Prerequisites

Run Benchmarks

Understanding Results

About

Releases

Packages

Used by

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

Caching libraries Benchmarks

Methodology

🖥️ System Information

📊 Set new (no eviction)

📊 Set at capacity (eviction)

📊 Lookup existing

📊 Lookup missing

📊 Deletion

📊 Popitem

📊 Hash collision

📊 Iterate items

🚀 Usage

Prerequisites

Run Benchmarks

Understanding Results

About

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages