3

I could need your experience according server implementations for the currently evolving Websockets in HTML5. We want to use this technology for a project using common webbrowsers on the client site. What we now need is a webserver that feeds our requirements. These are:

  • written in Java or PHP since these are the languages which currently are used and mastered by all the programmers in the project. Since it seems that there are no promising PHP implementations yet, I tend to concentrate on the Java site.
  • as fast as possible. We don't expect too heavy usage in production but we need to be able to synch two browsers on independent host in nearly real-time.
  • easy expandable. We will need to implement our own protocol on top of the websockets so we would like the most possible flexibility.
  • easy to use, well documented and actively developed. Since we have to come to a first running implementation rather quickly and since I expect some changes in the developing people during the time, it would be great if it's not to hard to get into the server programming.

I already had a look on some servers/frameworks which look promising, e.g. Jetty, Netty and jWebSocket. I also found some older discussions of this topic. However, I would be glad about some current experiences or recommendations of those of you who already have used Java (or PHP) to work with Websockets. Is there some implementation which is especially well-fitting or one that we absolutely shouldn't use?

Edit: For clarity: When I speak of implementing our own protocol on top of Websockets I mean having some kind of plugin mechanism so that we can use a Websocket to send XML-messages to the server and then use a plugin to interpret them into some server actions as we like. I guess every server implementation also supports this but it would be good if it was not too difficult to write such plugins.

Improve this question

1 Answer 1

Reset to default
1

I don't have any experience with PHP so I'll give you my opinion for the Java (and python side).

For Java:

You have your very common and safe choice of Tomcat. It's one of the most actively developed servers, its open source, lots of support/help from its very large community, from what I've read (but not tested) it can be fairly scalable and it also has a very mature console to monitor different metrics for your website.

You also have other choices such as Caucho's Resin, who have a very small yet smart team who actively develop their webserver. Their new heartbeat service allows a good degree of scalability (again from what I've read).

And then you also have Google's appengine which is all about scalability. Their API (for both, the server and client side) has started to focus on realtime web applications so it's a bit easier to for creating high performance web applications without having to worry about the server related optimization that you often have to make with other servers. They have APIs for creating "channels" between the server and client side for very realtime applications. The AppEngine team also has a new MapReduce API for analyzing data quickly/efficiently. For these reasons and the personal opinion that they are pioneering new web technologies/standards, I'm personally been leaning towards AppEngine. It's free if you're keeping cpu/data usuage low but if you need full scale deployment it does start to get pricey so do research. If your project is more for fun/school, its free quotas are enough and all its development tools/APIs are free so it's perfect.

For Python:

I'm more Java than Python, but Python is definitely a great language and very very very easy to learn. And the reason I'm recommending Python for consideration here is because AppEngine also allows developers the choice of Python. The author/creater of the Python language is now on the AppEngine team so, even though the Python and Java APIs are pretty much exactly the same, they sometimes have new/better APIs for Python. For example, I think the datastore API offered with Python is a bit easier and optimized for the Google datastore. Also, since you'd be using python, you can use Django with AppEngine.

The second choice that I've recently become aware of is the tornado web server which was developed by FriendFeed and then bought and used by Facebook. It's also actively developed. But the reason I bring this up is that Tornado's focus, much like AppEngine's, is high web application performance and scalability. I have no experience with it but I've become interested in its technology simply because no one can contest that Facebook requires a very high performance/scalable web server.

Some last thoughts:

In general you'll find that it isn't very difficult to build your own protocol on top of the APIs provided from most web servers (even Asp.NET MVC). But from what I've been able to test around with AppEngine, it really focuses on being flexible on what kind of protocol is build on top of it. To me, its pretty much just feels like a RPC server which which works on top of HTTP. And that becomes even more apparent if you use Python just because of some of the language constructs.

Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

Thanks for your answer. Googles AppEngine indeed looks promising but I fear it's not working with Websockets. Also we already have an existing server infrastructure (using Tomcat which also does not support Websockets unfortunately) and will us that for the project.
Yup, true. The issue has been Acknowledged by AppEngine and there are ways around this using TyphoonAE. Right now AppEngine has a Channel API for similar use case which google itself uses for things like Google Docs and GChat I believe. But I'm sure AppEngine will release the Web Sockets API soon as well.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.