eZ publish Evaluation: Version 2 Vs. Version 3

Flexibility

The flexibility of eZ publish versions 2 and 3 was compared in terms of the creation of custom content classes and page customization and layout. With regard to custom content classes, as mentioned above, eZ publish 3 allows users to identify their own content classes and attributes through the administrative back-end. In version 2 this required making manual changes to the database, coding new classes and functions and creating the matching front- and back-end templates.

In eZ publish, page customization and layout are separated from content, code, and multilingual functionality. Every page is generated dynamically, by means of a template. A template is one or more plain HTML files that contain template variables. These variables are filled in by eZ publish when a page is generated. In eZ publish 2, the template sets are spread among all modules. Each module has at least a subdirectory called ‘admin’ and usually a directory called ‘user’.

In eZ publish 3, all standard templates are stored in a single folder. The basic principle for creating custom designs in version 3 is to create within the custom design folders override templates that replace the standard templates. eZ publish 3 also introduces template operators and functions that allow advanced template customization.

Conclusion

The creation of new content classes and attributes is much easier in version 3 than it is in version 2. One of the advantages of Version 3 is that it allows users to identify their own content classes and attributes without coding or manual database updates. The only task that remains is to develop the necessary front-end templates for the new content class.

In version 3, the template system has become more advanced and flexible, but it’s also more cumbersome to implement. In comparison to version 2, the version 3 user needs to have a greater knowledge of the eZ publish application in order to be able to set up or change design templates. This could lead to a situation where an organization could become dependent on an external service provider for even minor changes to the look and feel of its site.

Performance

Scalability and performance are key factors in the evaluation of any Web platform. In this evaluation, eZ httpbench was used to gain an indication of the differences in the performance of eZ publish versions 2 and 3. The experiences of users as submitted to the forums on the eZ publish Website were also reviewed.

Using eZ httpbench, a tool developed by eZ Systems to benchmark HTTP pages, the average download times of pages with a size of approximately 12.0 KB were compared for the 2 versions. The following pages were fetched 25 times by the eZ httpbench benchmark tool (14.47 KB, eZ publish 2 and 11.78 KB, eZ publish 3). The results are shown in table 1 below. Unfortunately, eZ httpbench is no longer available as a download from the eZ Systems site, but it can still be downloaded here.

Note that these statistics are not typical for all installs of eZ publish. Throughput may vary according to server specifications, the overall server load, the configuration of the eZ publish application, and the use of accelerators. The results of a more extensive performance test of eZ publish version 3 are available on the eZ publish site.

Table 1. Comparison of eZ publish versions 2 and 3: average download times of pages with a size of approximately 12.0 KB

On the eZ publish Website’s developer forums, users report mixed experiences with regard to the performance of versions 2 and 3. Some users find that version 2 performs better, while others prefer version 3. The use of eZ publish caching and PHP caching applications has recently received a lot of attention, and useful suggestions have been made on how to use caching to improve performance.

According to one forum post from an eZ publish developer, ‘[…] we've just started with optimizations and you should see tremendous speed improvements from the 2.x series’. Another experienced user advises others to use ‘2 GHz dual cpu servers + php accelerator for a heavy load eZ publish 3 server’. Users report that both versions contain several poorly written SQL statements that slow down some parts of the application.

eZ Systems claims that eZ publish 3 is much more scalable than version 2. However, because of the flexibility of version 3, it is also easier to misconfigure the system, and thus experience poorer performance. eZ Systems has made this issue a focus point for the future development of version 3. We have not attempted to verify the claim of scalability in this discussion, but for more information on this topic, see this discussion.

Conclusion

From the eZ httpbench tests, it appears that eZ publish 3 is slightly slower than version 2. However, the performance of both versions 2 and 3 on a standard server is still more than adequate. The performance of both versions may be further improved by tuning the site with eZ publish caching enabled, or by using PHP caching applications, such as the ionCube PHP Accelerator, Zend, Alternativ PHP Cache or Turck MMCache.