C ++ vs Java for server application

Question

I need to make a server with a low latency, where raw data throughput is less important than the ability to handle thousands of concurrent connections. I do not know which language to use. Java is simpler, development will be faster, nice interface for everyone that I need + good support for networks (JAVA NIO and selector class). But in fact, I don’t have much experience with java other than programming, so even if I read things like java, today is not slower than my native C / C ++, there are still some doubts about performance, because c / C ++ is compiled into binary, but java.class files are interpreted, so there must be overhead in this translation.

• Can anyone post their own experience?
• Does someone have a low latency application programmed?
• What do you mean, is it better to use an application for this type?

EDIT: The application will be the server for real-time support in web applications (e.g. chat, etc.).

Answer 1

Usually people want low latency applications, so I don’t know what you mean by low latency.

so even if I read things like java, today is no slower than native C / C ++

Its still a little slower, just not enough to make the difference most of the time.

For example, a network IO is a network IO regardless of the language you use. for example, Language will not make your network faster or slower, except in cases of extremely low latency.

because c / c ++ is compiled into binary, but java.class files are interpreted,

Java compiles to native code at runtime, so in the end both systems run native code.

therefore, MUST be invoiced in this translation.

There is overhead at startup. But once the application is launched, the impact is much less.

Has someone programmed a high latency application?

If your network has a high latency, your language is less likely to matter.

What do you mean, is it better to use for this type of application?

I suggest you use the language that is most convenient for you.

The application will be a server for real-time support in web applications

In this case, your network latency will be much higher than your application if you encoded it correctly.

---

To give you a broad idea, a typical internet connection has a delay of 5-20 ms. A typical user can only respond with latency of 50 ms. A server with a delay of 20-50 ms can be written in almost any language. A server with a delay of 200 to 500 microseconds can be written in Java or C ++. If you need less than 100 microseconds, including latency for the client, you are looking at specialized equipment and some C ++ code for your main engine. In this case, the server and the consumer are likely to be in the same building, if not in the same rack.

Answer 2

Most JVM implementations use just-in-time (JIT) compilation methods, so they can be almost as effective as compiled programs (in C ++ or Ocaml).

The JVM usually translates to the machine code (in stages and once) the hot spots of your .class bytecode files.

You can use other languages ​​such as Ocaml (using Ocsigen for the web part) or Opa

Of course, you can always rewrite a very critical part of your application in C (or C ++) and call it from Java code (or Ocaml or another language).