Unscientific Jetty versus Glassfish for REST

This post was too unscientific and was updated. Jetty is an excellent container and the container of choice whenever I do something with servlets. Ever since we’ve developed SnipSnap some years ago I love Jetty. Glassfish has some very promising features like the admin console and I´m eager to try Glassfish in a project sometimes in the future.

Reading about another story of Rails performance, I grabbed JMeter to benchmark one of my current projects. Not so much as a comparison for Ruby – which managed 320 requests per second – but more as a comparison of Jetty and Glassfish.

The application is a small REST server which reads data from a JDBM storage, transforms it with my own framework to Json and delivers the result with Jersey.

Both servers were started with their default configuration through their maven plugins (wonderful easy to use mvn glassfish:run). Unscientific as it may be, the numbers are:

  • around 1000 requests/sec for both containers

Both with 200 threads and 50 requests per thread. Both numbers are great for my MacBookPro and good enough for me. They also are so close to each other so they are not a deciding factor for either Glassfish or Jetty. At the risk of comparing apples to oranges I have no fear of deploying this to a production system and scaling cheap (and even better with E-Tag caching), keeping in mind the requests per second with 25 servers in the Rails example.

Thanks for listening.

Update:: Another Rails application which thinks it did scale – at least with Merb, to 650k page views per day, well that’s “650K hits per day is ‘only’ around 8 per second (assumed a 20 hour day to spike it a little). This doesnt actual seem all that much?”.

The JMeter speed and 1000req/sec (for an admittantly simple REST GET) results in … 86.4M requests per day. Uh. On my MacBookPro.

REST: Lean JSON and XML from the same code

Generating JSON and XML with the same code is difficult. One can create the semantically richer XML and convert it to JSON, but JSON notations for XML like Badgerfish look quite ugly to JSON advocates.

The problem at the core is that XML is typed whereas JSON is not. Every node in XML needs a type – it’s name – for example <item><id>123</id><item>. JSON doesn’t need such a type, { id: 123 } is fine for an item. { item: {id: 123}} looks too verbose. Especially getting to the data in Javascript: var id = item.item.id. The same goes for accessing arrays with var id = items[0].item.id; instead of var id = items[0].id;. The problem exists with other dynamic languages and data structures too, see Cobra vs. Mongoose for Ruby.

As I currently develop a REST based Jersey application in Java I needed a way to generate lean JSON and XML. Wouldn’t it be best to have one code for both? DRY. My previous solution for generation JSON worked fine. The $(...) method calls create a node tree with nodes and lists. With a JsonRenderer and the Visitor pattern I generate JSON from the node tree. The problem was that this Java code

creates nice JSON like { id: 123, items: [ ... ] }, but was unable to generate XML. As written above, the outer list has no type and a XmlRender therefor cannot render <shoppinglist><id>123</id>...</shoppinglist>.

The solution I thought about is to add type information to nodes which have no names.

The implementation uses a simple static method and a Type class.

The type is attached to the node and if the node has no name but a type, the XmlRender uses the type instead of the name. The JsonRender doesn’t use the type information and renders the same JSON as before. The piece of Java code now generates XML

and lean JSON where neither shoppinglist nor item has a type

Next thing is to automatically apply the right renderer, toXml and toJson from within Jersey. The content negotiation then choses the accepted format for the client. Attributes (Meta-Information?) are not solved yet and I’m not sure if they are needed, or how to nicely add meta information to the $(...) tree. There is some discussion in the context of markup builders and attributes on James blog.

Probably the code will be released as an open source RESTkit if someone is interested.

Thanks for listening.

Problems with Jersey, REST, JSON and UTF-8 [Update]

UTF-8 is always a problem. Unbelievable. 2008 and we still haven’t fixed this. One of my current projects is a Javascript frontend with a REST backend. The backend stores to MySQL (a famous UTF-8 trouble maker) and creates JSON to REST calls. The problems starts with UTF-8 characters. Somewhere in the callchain – as always – characters don’t get correctly written. MySQL and the JDBC driver should work, the JSP page is UTF-8 (@page and meta-equiv), jQuery – which does the AJAX – and JS do know UTF-8 and Jersey should be UTF-8 too. But with some experiments now I’m quite sure that Jersey (JSR 311 REST framework) is to blame. I’m not sure how to specify UTF-8, this

doesn’t help. Funny, every major project with several frameworks along the call chain and several languages (JS, C, Java) makes UTF-8 problems somehow. I’m so fed up with this, it’s 2008.

Update: Jersey uses InputStreams for all encodings, especially StringProvider is relevant to me (se above). Does this work with Unicode?