refactor: improve click buffer performance

This commit is contained in:
sjdonado
2025-11-02 08:48:10 +01:00
parent 046b15bdce
commit 0c89fad713
5 changed files with 236 additions and 146 deletions
+3 -3
View File
@@ -2,9 +2,9 @@
[![Docker Stars](https://img.shields.io/docker/stars/sjdonado/bit.svg)](https://hub.docker.com/r/sjdonado/bit)
[![Docker Image Size](https://img.shields.io/docker/image-size/sjdonado/bit/latest)](https://hub.docker.com/r/sjdonado/bit)
Lightweight URL shortener (API-only) with minimal resource requirements. Avg memory consumption **30MiB**, avg CPU load 20%.
Lightweight URL shortener (API-only) with minimal resource requirements.
Highly performant: **7.9k req/sec**, latency 15.8ms (100k requests using 125 connections, [benchmark](docs/SETUP.md#benchmark)).
Highly performant: **11k req/sec**, latency 11ms, 40MiB avg memory usage (100k requests using 125 connections, [benchmark](docs/SETUP.md#benchmark)).
Self-hosted: [Dokku](docs/SETUP.md#dokku), [Docker Compose](docs/SETUP.md#docker-compose).
@@ -20,7 +20,7 @@ It is feature-complete by design: simple and reliable without unnecessary bloat.
## Recommented requirements
- 100MB disk space
- 70MiB RAM
- 100MiB RAM
- x86_64 or ARM64
## Documentation
+19 -95
View File
@@ -4,90 +4,6 @@ module App::Controllers
include App::Lib
include App::Services
# Buffered channel to hold click data
@@click_channel = Channel(NamedTuple(
link_id: Int64,
remote_address: String,
user_agent: String?,
referer: String
)).new(1024)
@@processor_started = begin
spawn do
batch = [] of NamedTuple(
link_id: Int64,
remote_address: String,
user_agent: String?,
referer: String
)
batch_size = 32
min_batch = 32
max_batch = 512
scaling_factor = 1.5
loop do
select
when click_data = @@click_channel.receive
batch << click_data
if batch.size >= batch_size
process_click_batch(batch)
batch.clear
# Increase batch size when reaching capacity
batch_size = (batch_size * scaling_factor).to_i.clamp(min_batch, max_batch)
end
when timeout(1.seconds)
unless batch.empty?
process_click_batch(batch)
batch.clear
batch_size = (batch_size / scaling_factor).to_i.clamp(min_batch, max_batch)
else
# Reset to default if idle
batch_size = min_batch unless batch_size == min_batch
end
end
end
end
true
end
private def self.process_click_batch(batch)
clicks = [] of App::Models::Click
batch.each do |click_data|
begin
client_ip = IpLookup.ip_from_address(click_data[:remote_address])
family, _, _, os = UserAgent.parse(click_data[:user_agent] || "")
click = App::Models::Click.new
click.link_id = click_data[:link_id]
click.country = client_ip ? IpLookup.country(client_ip) : nil
click.user_agent = click_data[:user_agent]
click.browser = family
click.os = os.try &.[0] # OS family
click.referer = click_data[:referer]
clicks << click
rescue ex
Log.error { "Click data processing error: #{ex.message}" }
end
end
# Batch insert clicks if any were successfully processed
unless clicks.empty?
begin
multi = Crecto::Multi.new
clicks.each do |click|
multi.insert(click)
end
Database.transaction(multi)
rescue ex
Log.error { "Batch click insertion error: #{ex.message}" }
end
end
end
def self.redirect_handler
->(env : HTTP::Server::Context) {
link_id, url = Database.raw_query("SELECT id, url FROM links WHERE slug = (?) LIMIT 1", env.params.url["slug"]) do |result|
@@ -96,21 +12,29 @@ module App::Controllers
remote_address = env.request.headers["Cf-Connecting-Ip"]? || env.request.remote_address.to_s
# Send redirect immediately
env.response.status_code = 301
env.response.headers.add("Location", url)
env.response.headers.add("X-Forwarded-For", remote_address)
begin
@@click_channel.send({
link_id: link_id,
remote_address: remote_address,
user_agent: env.request.headers["User-Agent"]?,
referer: env.request.headers["Referer"]?.try { |r| URI.parse(r).host rescue r } || env.params.query["utm_source"]? || "Direct"
})
rescue Channel::ClosedError
Log.error { "Click channel closed" }
rescue ex
Log.error { "Error queuing click: #{ex.message}" }
# non-blocking click proccessing
spawn do
begin
client_ip = IpLookup.ip_from_address(remote_address)
family, _, _, os = UserAgent.parse(env.request.headers["User-Agent"]? || "")
click = App::Models::Click.new
click.link_id = link_id
click.country = client_ip ? IpLookup.country(client_ip) : nil
click.user_agent = env.request.headers["User-Agent"]?
click.browser = family
click.os = os.try &.[0]
click.referer = env.request.headers["Referer"]?.try { |r| URI.parse(r).host rescue r } || env.params.query["utm_source"]? || "Direct"
Database.insert(click)
rescue ex
Log.error { "Click tracking error: #{ex.message}" }
end
end
}
end
+211 -45
View File
@@ -73,7 +73,7 @@ Recommended for lower latency communication (no host network traversal)
## Local Development
### Requirements
- Crystal 1.12+
- Crystal 1.18+
- Shards package manager
- SQLite3
@@ -109,59 +109,225 @@ ENV=test crystal spec
## Benchmark
- Colima: cpu 1, mem 1
- SoC: Apple M3 Pro
### Run
```
~/p/bit> colima start --cpu 1 --memory 1
INFO[0000] starting colima
INFO[0000] runtime: docker
INFO[0001] starting ... context=vm
INFO[0076] provisioning ... context=docker
INFO[0077] starting ... context=docker
INFO[0077] done
~/p/bit> ./benchmark.cr
Setting up...
Waiting for the application to be ready...
Seeding the database...
Checking seed results...
Fetching all created links from /api/links...
Selected link for benchmarking: http://localhost:4000/slug2576
Starting benchmark with Bombardier...
Bombarding http://localhost:4000/slug2576 with 100000 request(s) using 125 connection(s)
100000 / 100000 [===================================================================================================================================================================] 100.00% 7795/s 12s
shards build --release --no-debug --progress --stats
shards run benchmark
```
### Output
Chip: Apple M4 Pro. Memory: 24GB
- Dry run
```
1762068811 ~/p/bit> shards build --release --no-debug --progress --stats
Dependencies are satisfied
Building: bit
Parse: 00:00:00.000041417 ( 1.17MB)
Semantic (top level): 00:00:00.407816208 ( 163.45MB)
Semantic (new): 00:00:00.001814125 ( 163.45MB)
Semantic (type declarations): 00:00:00.019943333 ( 179.45MB)
Semantic (abstract def check): 00:00:00.007606542 ( 195.45MB)
Semantic (restrictions augmenter): 00:00:00.006390917 ( 195.45MB)
Semantic (ivars initializers): 00:00:00.012892709 ( 211.45MB)
Semantic (cvars initializers): 00:00:00.101427125 ( 211.50MB)
Semantic (main): 00:00:00.639027292 ( 499.88MB)
Semantic (cleanup): 00:00:00.000513000 ( 499.88MB)
Semantic (recursive struct check): 00:00:00.000625542 ( 499.88MB)
Codegen (crystal): 00:00:00.514694833 ( 532.38MB)
Codegen (bc+obj): 00:00:14.734378459 ( 532.38MB)
Codegen (linking): 00:00:00.308837000 ( 532.38MB)
Macro runs:
- /opt/homebrew/Cellar/crystal/1.18.2/share/crystal/src/ecr/process.cr: reused previous compilation (00:00:00.003361958)
Codegen (bc+obj):
- no previous .o files were reused
Building: cli
Parse: 00:00:00.000057625 ( 1.17MB)
Semantic (top level): 00:00:00.378950750 ( 163.45MB)
Semantic (new): 00:00:00.001392542 ( 163.45MB)
Semantic (type declarations): 00:00:00.017725458 ( 179.45MB)
Semantic (abstract def check): 00:00:00.007331291 ( 195.45MB)
Semantic (restrictions augmenter): 00:00:00.006174250 ( 195.45MB)
Semantic (ivars initializers): 00:00:00.012456209 ( 211.45MB)
Semantic (cvars initializers): 00:00:00.101925250 ( 211.50MB)
Semantic (main): 00:00:00.283259791 ( 371.62MB)
Semantic (cleanup): 00:00:00.000385375 ( 371.62MB)
Semantic (recursive struct check): 00:00:00.000574250 ( 371.62MB)
Codegen (crystal): 00:00:00.318639083 ( 387.88MB)
Codegen (bc+obj): 00:00:00.090703209 ( 387.88MB)
Codegen (linking): 00:00:00.100725000 ( 387.88MB)
Codegen (bc+obj):
- all previous .o files were reused
Building: benchmark
Parse: 00:00:00.000210708 ( 1.17MB)
Semantic (top level): 00:00:00.259058375 ( 147.78MB)
Semantic (new): 00:00:00.000878709 ( 147.78MB)
Semantic (type declarations): 00:00:00.012123625 ( 147.78MB)
Semantic (abstract def check): 00:00:00.032016792 ( 147.78MB)
Semantic (restrictions augmenter): 00:00:00.004018334 ( 147.78MB)
Semantic (ivars initializers): 00:00:00.006835041 ( 147.78MB)
Semantic (cvars initializers): 00:00:00.031038959 ( 195.78MB)
Semantic (main): 00:00:00.157428625 ( 243.83MB)
Semantic (cleanup): 00:00:00.000264416 ( 243.83MB)
Semantic (recursive struct check): 00:00:00.000380166 ( 243.83MB)
Codegen (crystal): 00:00:00.079188459 ( 259.83MB)
Codegen (bc+obj): 00:00:04.807389083 ( 259.83MB)
Codegen (linking): 00:00:00.098161042 ( 259.83MB)
Codegen (bc+obj):
- no previous .o files were reused
1762068874 ~/p/bit> shards run benchmark
Dependencies are satisfied
Building: benchmark
Executing: benchmark
Starting application: ./bit...
Application output will be saved to: app_output.log
Application started with PID: 12693
Checking if server is ready at http://localhost:4000...
.Server is ready!
Seeding database...
Database seeded successfully.
Fetching links from API...
Selected link: http://localhost:4000/slug9623
Starting benchmark with 100000 requests...
Bombarding http://localhost:4000/slug9623 with 100000 request(s) using 125 connection(s)
100000 / 100000 [===============================================================================================================================================================================================] 100.00% 11079/s 9s
Done!
Statistics Avg Stdev Max
Reqs/sec 7900.70 7570.15 29263.59
Latency 15.89ms 10.22ms 67.32ms
Reqs/sec 11361.35 8907.62 28610.94
Latency 11.09ms 6.66ms 52.76ms
Latency Distribution
50% 4.82ms
75% 9.24ms
90% 51.61ms
95% 52.74ms
99% 55.07ms
50% 1.93ms
75% 3.12ms
90% 39.47ms
95% 40.07ms
99% 42.60ms
HTTP codes:
1xx - 0, 2xx - 0, 3xx - 100000, 4xx - 0, 5xx - 0
others - 0
Throughput: 2.14MB/s
Throughput: 3.06MB/s
Benchmark completed successfully.
Analyzing resource usage...
Timestamp CPU(%) Memory(MiB)
1742763202 0.01 44.83
1742763204 0.01 44.78
1742763206 91.53 68.23
1742763208 92.03 68.17
1742763210 91.0 68.09
1742763212 92.73 68.38
1742763214 92.17 67.66
1742763216 91.1 67.69
1742763218 2.93 67.04
**** Resource Usage Statistics ****
Measurements: 9
Average CPU Usage: 61.5%
Average Memory Usage: 62.76 MiB
Peak CPU Usage: 92.73%
Peak Memory Usage: 68.38 MiB
Cleanup completed. Resource usage data saved in resource_usage.txt
Measurements: 12
Average CPU Usage: 73.94%
Average Memory Usage: 42.45 MiB
Peak CPU Usage: 100.0%
Peak Memory Usage: 56.02 MiB
**** Files Generated ****
Resource stats: resource_usage.log
Application log: app_output.log
Stopping application...
Application stopped.
```
- Second run
```
1762068874 ~/p/bit> shards run benchmark
Dependencies are satisfied
Building: benchmark
Executing: benchmark
Starting application: ./bit...
Application output will be saved to: app_output.log
Application started with PID: 12693
Checking if server is ready at http://localhost:4000...
.Server is ready!
Seeding database...
Database seeded successfully.
Fetching links from API...
Selected link: http://localhost:4000/slug9623
Starting benchmark with 100000 requests...
Bombarding http://localhost:4000/slug9623 with 100000 request(s) using 125 connection(s)
100000 / 100000 [===============================================================================================================================================================================================] 100.00% 11079/s 9s
Done!
Statistics Avg Stdev Max
Reqs/sec 11361.35 8907.62 28610.94
Latency 11.09ms 6.66ms 52.76ms
Latency Distribution
50% 1.93ms
75% 3.12ms
90% 39.47ms
95% 40.07ms
99% 42.60ms
HTTP codes:
1xx - 0, 2xx - 0, 3xx - 100000, 4xx - 0, 5xx - 0
others - 0
Throughput: 3.06MB/s
Benchmark completed successfully.
**** Resource Usage Statistics ****
Measurements: 12
Average CPU Usage: 73.94%
Average Memory Usage: 42.45 MiB
Peak CPU Usage: 100.0%
Peak Memory Usage: 56.02 MiB
**** Files Generated ****
Resource stats: resource_usage.log
Application log: app_output.log
Stopping application...
Application stopped.
1762068900 ~/p/bit> shards run benchmark
Dependencies are satisfied
Building: benchmark
Executing: benchmark
Starting application: ./bit...
Application output will be saved to: app_output.log
Application started with PID: 18421
Checking if server is ready at http://localhost:4000...
.Server is ready!
Seeding database...
Runtime error near line 1: UNIQUE constraint failed: users.api_key (19)
Runtime error near line 7: UNIQUE constraint failed: links.slug (19)
Warning: Database seeding failed. Continuing anyway...
Fetching links from API...
Selected link: http://localhost:4000/slug5911
Starting benchmark with 100000 requests...
Bombarding http://localhost:4000/slug5911 with 100000 request(s) using 125 connection(s)
100000 / 100000 [===============================================================================================================================================================================================] 100.00% 11080/s 9s
Done!
Statistics Avg Stdev Max
Reqs/sec 11213.27 8842.12 29594.78
Latency 11.24ms 6.87ms 43.66ms
Latency Distribution
50% 1.96ms
75% 3.01ms
90% 40.02ms
95% 41.06ms
99% 42.58ms
HTTP codes:
1xx - 0, 2xx - 0, 3xx - 100000, 4xx - 0, 5xx - 0
others - 0
Throughput: 3.02MB/s
Benchmark completed successfully.
**** Resource Usage Statistics ****
Measurements: 12
Average CPU Usage: 74.28%
Average Memory Usage: 32.18 MiB
Peak CPU Usage: 100.0%
Peak Memory Usage: 40.88 MiB
**** Files Generated ****
Resource stats: resource_usage.log
Application log: app_output.log
Stopping application...
Application stopped.
```
+1 -1
View File
@@ -21,7 +21,7 @@ OptionParser.parse do |parser|
parser.on("--update-parsers", "Download UA regexes and/or GeoLite2 database") do
puts "=== Starting data files update ==="
App::Services::Cli.update_uap_regexes
App::Services::Cli.download_geolite_db
App::Services::Cli.update_geolite_db
puts "=== Data files updated successfully ==="
exit
end
+2 -2
View File
@@ -1,5 +1,5 @@
name: bit
version: 1.5.3
version: 1.6.0
authors:
- Juan Rodriguez Ronado <@sjdonado>
@@ -32,6 +32,6 @@ development_dependencies:
spec-kemal:
github: kemalcr/spec-kemal
crystal: ">= 1.12.1"
crystal: ">= 1.18.2"
license: MIT