Macromedia Flex Performance Brief

by kai on 06/06/2004



Macromedia delivered an article on their DevNet which gives an impression of how a MXML-based Flex application performs again a JSP-based HTML application using the same backend application tier.

The results are really interesting. In terms of memory usage there’s no difference – cause both approaches are using a JRun instance – and regarding the network load they notice a light advantage for the MXML app. But: When we talk about average CPU load using a load test tool like Microsofts web stress tool or something similar, Flex beats the JSP approach for miles… The main reason for this behaviour is that Flex doesn’t have to open and close socket connections all the time (several for each and every “request” to the server) because the application resides on the client and just goes back for the server to fetch or send data.

I’d really love to have a look on the JSP code they used to compare with the MXML application and how well optimized it was. Also interesting: The script files for the stress simulation on the MXML application. How does that work with the Microsoft tool, I never used that one strongly. Up to now my favourite was the Paessler Webserver Stress Tool.

Performance Brief (pdf)

Paul June 6, 2004 at 12:00 am

I think the jsp solution is not realy comparable to the felx solution when it comes to “caching” or reloading of redundant information:
“Flex applications are delivered as complete, self-contained applications, while JSP applications generally require a round trip to the server to generate the next page of the application.”
as far as I understood they implmented the whole jsp-shopping solution in a singlepage-html-table-layout, not in a frameset. Therefore the have tons of reloads for every action that is taking place. The felx solution can of course skip this relaoding things, since only the new informations come from the server, while the allready loaded articles for example are still present in the application. Even with an optimized jsp solution the cpu usage might be higher, but the difference shouldn’t be that big.

As you said, it would be interesting to see JSP code used.

Regards,
Paul

Comments on this entry are closed.

Previous post:

Next post: