Repository navigation
Benchmark
Levtus is optimized for extreme throughput and minimal resource footprint by combining Project Loom (Virtual Threads) with GraalVM Native Image technology.
The following "Hello World" application was used for all tests. It includes standard DoS protection settings and static file configuration.
import io.github.bernardusz.levtus.Levtus;
void main(){
Levtus app = Levtus.create();
app.maxConcurrentConnections(500)
.maxBodySize(1024 * 1024)
.maxHeaderCount(50)
.maxHeaderSize(4096)
.maxLineSize(4096)
.maxEmptyLines(3)
.staticFiles("public");
app.get("/", ctx -> {
ctx.text("An incoming " + ctx.req().method() + "request!" );
});
app.post("/", ctx -> {
ctx.text(ctx.req().body().toString());
ctx.status(200).contentType("text/plain").send("Created!");
});
app.put("/", ctx -> {
ctx.text(ctx.req().body().toString());
});
app.delete("/", ctx -> {
ctx.text("Deleted successfully!");
});
app.listen(8080);
}| Metric | JVM (Fat Jar) | GraalVM Native |
|---|---|---|
| Throughput (RPS) | ~209,223 | ~160,998 |
| Idle RAM | ~45.6 MB | 11-13 MB |
| Peak RAM | ~310 MB | 114 MB |
| Binary Size | ~38 KB (jar) | ~16.7 MB |
Best for: High-density microservices, Serverless, and CLI tools.
- Binary Size: 16.71 MB
- Build Time: 1m 5s
- Peak Build RSS: 1.78 GB
Tested with 4 threads and 400 concurrent connections over 30 seconds.
Requests/sec: 160,998.28
Transfer/sec: 16.74 MB
Latency Distribution:
50% 1.68ms
75% 4.04ms
90% 6.55ms
99% 15.28ms
Max 98.06ms
- Idle: ~12.0 MB
- Peak: ~114 MB
Idle footprint at ~12MB
Best for: High-traffic core services requiring maximum throughput.
Tested with 4 threads and 400 concurrent connections over 30 seconds.
Requests/sec: 209,223.55
Transfer/sec: 21.75 MB
Latency Distribution:
50% 1.04ms
75% 3.07ms
90% 5.30ms
99% 12.11ms
Max 206.83ms
- Idle: ~45.6 MB
- Peak: ~310 MB
Idle footprint at ~45MB
- Go-Level Efficiency: In Native Image mode, Levtus achieves an idle RAM footprint of 11-13 MB, putting it on par with Go for resource efficiency while maintaining the power of the Java ecosystem.
- Loom Scaling: Handling 400 concurrent connections with a median latency as low as 1.04ms demonstrates the lightweight nature of Virtual Threads compared to traditional thread pools.
-
Deployment Flexibility:
- Native Image: Provides near-instant startup (<10ms) and minimal RAM, perfect for containerized environments.
- JVM Mode: Provides higher raw throughput (~200k+ RPS) by utilizing the Graal JIT compiler's advanced runtime optimizations.
bernardus@DESKTOP-3HNT5JD:~/projects/web/levtus.wiki/levtus-test/target$ native-image -jar levtus-test-1.0-SNAPSHOT.jar
========================================================================================================================
GraalVM Native Image: Generating 'levtus-test-1.0-SNAPSHOT' (executable)...
========================================================================================================================
[1/8] Initializing... (5.5s @ 0.12GB)
Java version: 25.0.3+9-LTS, vendor version: Oracle GraalVM 25.0.3+9.1
Graal compiler: optimization level: 2, target machine: x86-64-v3, PGO: ML-inferred
C compiler: gcc (linux, x86_64, 13.3.0)
Garbage collector: Serial GC (max heap size: 80% of RAM)
1 user-specific feature(s):
- com.oracle.svm.thirdparty.gson.GsonFeature
------------------------------------------------------------------------------------------------------------------------
Build resources:
- 6.27GB of memory (75.6% of system memory, less than 8GB of memory available)
- 8 thread(s) (100.0% of 8 available processor(s), determined at start)
[2/8] Performing analysis... [*****] (13.6s @ 0.49GB)
3,694 types, 4,250 fields, and 19,620 methods found reachable
1,191 types, 40 fields, and 1,098 methods registered for reflection
59 types, 59 fields, and 53 methods registered for JNI access
0 downcalls and 0 upcalls registered for foreign access
4 native libraries: dl, pthread, rt, z
[3/8] Building universe... (1.7s @ 0.54GB)
[4/8] Parsing methods... [**] (3.9s @ 0.59GB)
[5/8] Inlining methods... [***] (0.9s @ 0.63GB)
[6/8] Compiling methods... [******] (33.6s @ 0.60GB)
[7/8] Laying out methods... [**] (2.5s @ 0.70GB)
[8/8] Creating image... [**] (2.3s @ 0.78GB)
7.15MB (40.31%) for code area: 10,302 compilation units
8.52MB (48.00%) for image heap: 135,570 objects and 55 resources
2.07MB (11.69%) for other data
17.75MB in total image size, 16.71MB in total file size
------------------------------------------------------------------------------------------------------------------------
Top 10 origins of code area: Top 10 object types in image heap:
4.86MB java.base 1.91MB byte[] for code metadata
1.66MB svm.jar (Native Image) 1.40MB byte[] for java.lang.String
116.00kB java.logging 820.10kB java.lang.String
69.05kB levtus-test-1.0-SNAPSHOT.jar 472.94kB heap alignment
59.77kB com.oracle.svm.svm_enterprise 460.86kB byte[] for general heap data
48.81kB org.graalvm.nativeimage.configure 452.54kB java.lang.Class
40.29kB jdk.graal.compiler 384.18kB com.oracle.svm.core.hub.DynamicHubCompanion
36.69kB org.graalvm.nativeimage.base 361.34kB java.util.concurrent.ConcurrentHashMap$Node
21.62kB jdk.internal.vm.ci 268.70kB java.util.HashMap$Node
21.14kB jdk.net 250.07kB java.lang.Object[]
62.59kB for 8 more packages 1.74MB for 1071 more object types
Use '--emit build-report' to create a report with more details.
------------------------------------------------------------------------------------------------------------------------
Security report:
- Binary includes Java deserialization.
- CycloneDX SBOM with 7 component(s) is embedded in binary (519B).
- Advanced obfuscation not enabled; enable with '-H:AdvancedObfuscation=""' (experimental support).
------------------------------------------------------------------------------------------------------------------------
Recommendations:
G1GC: Use the G1 GC ('--gc=G1') for improved latency and throughput.
PGO: Use Profile-Guided Optimizations ('--pgo') for improved throughput.
FUTR: Use '--future-defaults=all' to prepare for future releases.
HEAP: Set max heap for improved and more predictable memory usage.
CPU: Enable more CPU features with '-march=native' for improved performance.
------------------------------------------------------------------------------------------------------------------------
5.0s (7.5% of total time) in 1208 GCs | Peak RSS: 1.78GB | CPU load: 6.47
------------------------------------------------------------------------------------------------------------------------
Build artifacts:
/home/bernardus/projects/web/levtus.wiki/levtus-test/target/levtus-test-1.0-SNAPSHOT (executable)
========================================================================================================================
Finished generating 'levtus-test-1.0-SNAPSHOT' in 1m 5s.bernardus@DESKTOP-3HNT5JD:~/projects/web/levtus.wiki/levtus-test/target$ ls -lh
total 16M
drwxr-xr-x 2 bernardus bernardus 4.0K Jun 9 17:18 classes
drwxr-xr-x 3 bernardus bernardus 4.0K Jun 9 17:18 generated-sources
-rwxr-xr-x 1 bernardus bernardus 16M Jun 9 17:22 levtus-test-1.0-SNAPSHOT -> Graal VM Binary
-rw-r--r-- 1 bernardus bernardus 38K Jun 9 17:19 levtus-test-1.0-SNAPSHOT.jar -> Fat Jar
drwxr-xr-x 2 bernardus bernardus 4.0K Jun 9 17:18 maven-archiver
drwxr-xr-x 3 bernardus bernardus 4.0K Jun 9 17:18 maven-status
-rw-r--r-- 1 bernardus bernardus 3.3K Jun 9 17:18 original-levtus-test-1.0-SNAPSHOT.jarbernardus@DESKTOP-3HNT5JD:~/projects/web/levtus.wiki/levtus-test/target$ ./levtus-test-1.0-SNAPSHOT
🚀 Levtus Engine started on port 8080 (HTTP) 28729 bernardus 20 0 32.1G 12012 10744 S 0.0 0.1 0:00.00 ./levtus-test-1.0-SNAPSHOT
bernardus@DESKTOP-3HNT5JD:~/projects/web/levtus.wiki/levtus-test/target$ wrk -t4 -c400 -d30s --latency http://localhost:8080
Running 30s test @ http://localhost:8080
4 threads and 400 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 2.93ms 3.98ms 98.06ms 91.10%
Req/Sec 40.54k 11.46k 88.99k 75.08%
Latency Distribution
50% 1.68ms
75% 4.04ms
90% 6.55ms
99% 15.28ms
4843143 requests in 30.08s, 503.45MB read
Requests/sec: 160998.28
Transfer/sec: 16.74MB
bernardus@DESKTOP-3HNT5JD:~/projects/web/levtus.wiki/levtus-test$ java -jar target/levtus-test-1.0-SNAPSHOT.jar
🚀 Levtus Engine started on port 8080 (HTTP) 23906 bernardus 20 0 4688M 45624 26316 S 0.0 0.6 0:00.00 java -jar target/levtus-test-1.0-SNAPSHOT.jar
bernardus@DESKTOP-3HNT5JD:~/projects/web/levtus.wiki$ wrk -t4 -c400 -d30s --latency http://localhost:8080
Running 30s test @ http://localhost:8080
4 threads and 400 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 2.18ms 2.72ms 206.83ms 88.14%
Req/Sec 52.75k 16.45k 124.01k 73.74%
Latency Distribution
50% 1.04ms
75% 3.07ms
90% 5.30ms
99% 12.11ms
6292708 requests in 30.08s, 654.13MB read
Requests/sec: 209223.55
Transfer/sec: 21.75MB
Video for proof of benchmark coming soon.