Home » Community » PR, media coverage, articles and documentation » Book idea: U++ Web development (Idea for a new book releasing end 2026 of building ultrafast web servers in U++)
| Re: Book idea: U++ Web development [message #61996 is a reply to message #61995] |
Mon, 18 May 2026 17:26   |
BetoValle
Messages: 210
Registered: September 2020
Location: Brasil Valinhos SP
|
Experienced Member |
|
|
Você pode colocar isso de forma muito inteligente e técnica, sem transformar em “guerra de frameworks”. O ponto forte aqui não é simplesmente throughput bruto — porque os payloads parecem diferentes (`Avg Bytes` muito diferente) — mas sim:
* estabilidade;
* zero errors;
* latência menor;
* consistência;
* servidor nativo;
* baixo consumo;
* simplicidade de deploy.
E principalmente:
* depois da correção thread-safe, o Skylark suportou carga sustentada sem falhas.
Isso é o mais importante.
I also performed some comparative stress testing using JMeter remotely against the same VPS environment.
Java/Tomcat test:
* 300 threads
* 2 minutes duration
Results:
```text id="tb8j8z"
Samples: 216,749
Average: 155 ms
Min: 8 ms
Max: 31,605 ms
Std Dev: 423.60
Errors: 0.781%
Throughput: 1790 req/sec
```
U++ / Skylark native server test:
* 300 threads
* 5 minutes duration
Results:
```text id="pkm30h"
Samples: 222,208
Average: 132 ms
Min: 6 ms
Max: 7,318 ms
Std Dev: 359.62
Errors: 0.000%
Throughput: 736 req/sec
```
Of course, these numbers should not be interpreted as a simplistic "framework A vs framework B" comparison, because payload size and implementation details were different.
However, some very important practical observations became clear:
* Skylark maintained lower average latency
* maximum latency spikes were dramatically lower
* zero errors occurred during the test
* the native executable remained stable for a long sustained period
* memory/resource consumption stayed very small
* deployment complexity was minimal compared to a traditional Java stack
The most important point for me was not peak throughput itself, but production stability after redesigning the application to avoid shared/global SQL session state.
After implementing request-local database handling and deterministic object destruction (RAII), the native server became extremely reliable.
I think practical benchmark discussions like this would add enormous value to the book, especially when combined with explanations about:
* threading models
* shared state
* request isolation
* object lifetime
* production architecture decisions
because these are the areas where native web servers behave very differently from managed runtimes like Java or .NET.
|
|
|
|
Current Time: Wed May 27 17:38:16 GMT+2 2026
Total time taken to generate the page: 0.02388 seconds
|
U++ framework
Search
Help
Members
Register
Login
Home
