Check out Olio over at Apache.org; it’s a sample Web app, three implementations, a load driver, and an invitation to build more. So all of you who are talking up your hot new Web frameworks, about how they’re simpler than Rails and faster than Java: Prove it. On a level playing field. What a great idea.
Contributions
Comment feed for ongoing:
From: Avi Bryant (Nov 14 2008, at 12:45)
I like that the spec here doesn't say anything about the URL structure of the application, which can legitimately differ between web frameworks.
Unfortunately, it looks like the benchmarking scripts have specific assumptions about URLs and query parameters hardcoded in, and there's a separate script for each implementation. I'd love to see this done with a universal load script that simulated an actual user going through the app, and followed links or submitted forms based on id attributes, rather than one which submits to predefined URLs.
[link]
From: tw3k (Nov 14 2008, at 13:29)
Tim, I have a question I've been meaning to ask you but only vaguely related to the topic of your posted.
I've been planing on writing a simple web app that uses the metadata of image files as a sort of flatfile db.
I am an artist and am incredibly bad at documenting my work. Once in a while I get around to taking a pictures of the work and organizing collections of images but inevitably I change disks/OS/server/software and end up losing a good amount of the data associated with the images.
Ultimately, I just want to write a simple web app that will allow me to easily read/write the metadata of the images and organize collections.
Given your interest in photography and programing I thought you might have some
good insight and be a good person to ask about this sort of web app.
[link]
From: Rob Sayre (Nov 14 2008, at 14:04)
Back in the Web 1.0 days, there was the J2EE Pet Store.
Didn't turn out to be such a great idea.
[link]
From: DanF (Nov 14 2008, at 17:40)
Cool idea. I love the stuff Sun is doing; I hope you guys come through this downturn alright.
I've been working feverishly on a Ruby on Rails project this week (should be working on it now...) and I've learned two things:
1. Netbean's is nice for RoR. I was using emacs, but some of the project navigation features in Netbeans speed things up a lot when you have ridiculous amounts of files in play.
2. I hate rails.
I think rails is almost enough to put me off Ruby -- it seems to be the paragon example of a DSL gone horribly wrong. Of course that's not Ruby's fault.
[link]
From: Akara Sucharitakul (Nov 24 2008, at 12:45)
A response to Avi Bryant's post at the top of these comments:
First of all, thank you very much for your comments on Olio.
As for the workload driver, you're right, it does have many URLs and parameters hardcoded - but not all. The Olio team is planning to merge the driver for each of the implementations into one universal driver handling variations in the URL. However, even the current load script does simulate actual users going through the application, and following links as you suggested. Moreover, the decision how to go through the application is driven by a stochastic matrix. This provides controlled randomness of the path each user clicks through.
The main reason the Olio workload driver does not parse more html than necessary is to reduce the overhead of the workload itself, thus allowing it to drive large scale tests with as little resources as possible.
Please do post ideas and responses on olio-dev@incubator.apache.org. The team is taking your input seriously.
[link]