Update Performance of Yii

[size="3"]What about to update this information ? [/size]

[size="3"][color="#4D90F0"][font="arial, sans-serif"]This schedule[/font][/color][color="#333333"][font="arial, sans-serif"] [/font][/color][color="#333333"][font="arial, sans-serif"]no longer[/font][/color][color="#333333"][font="arial, sans-serif"] [/font][/color][color="#333333"][font="arial, sans-serif"]relevant [/font][/color][/size]

We have source code for this benchmark http://code.google.com/p/phpmark/ but, unfortunitely not enough time to update it. Probably will move it to github to allow pull-requests.

I mentioned it on github already: There is a comparison that is only slightly out-of-date. I think it could be brought into proper shape really quickly by the community. As for the benchmarking: I’m preparing a quadcore machine for some benchmarking with Fedora 17, so I could generate some initial data. it’s not server-grade hardware (i.e. there is no 15,000RPM HDD and it’s limited to 8GB of RAM), but I think the results could be representative.

[size="3"] Da:Sourcerer, It wiil be great, if you do this[/size]

Meh. This is going to take a while.

I start to understand why there are so few benchmarks. It really is a lot of work and I’m still far from complete :( This really requires to read the documentation for most frameworks and create a small application for them as they provide no “Hello World” demo.

I managed to draw a few frameworks together and produce this (inconclusive) benchmark:

2981

benchmark.png

This has been conducted on a notebook equipped with a AMD E-450 CPU/GPU unit running Fedora 17 with a kernel patched and optimized for interactivity - which might explain some oddeties. E.g. Laravel is impressive but hardly that much ;D

What a strange image :)

Although Yii performances are good, from a marketing point of view I wouldn’t include a benchmark where Yii doesn’t win.

Something is wrong there…

I blame it on a non-optimized Apache/PHP environment and a custom kernel. Perhaps I should’ve booted into the stock kernel but I’ve been too impatient to get a quick glance ::)

By the way: samdark, are there any plans to revive the phpmark project? I’m a long way from being able to provide a really complete (i.e. go beyond simple “Hello World” demos) and fair (i.e. provide automated stats on memory consumption, disk I/O, etc) benchmarks. There will also be quite a lot of administrative overhead. The power of git submodules eases that a bit. But still …

Hm, the original phpmark has been set in place to provide a fair comparison. But from a marketing point, you’re actualy right. In the full benchmark, I’d like to include metrics like the spread of response times. As I see it, Yii is pretty constant there …

See above ;D

Suggestion: don’t use Fedora but use a - clean - server-grade distro (i.e. Ubuntu Server LTS, RHEL, CentOS, Scientific Linux).

Fedora is my choice for desktop systems. Everything else is on CentOS/RHEL, so don’t worry. This has just been a quick peek.

Edit: Coming to think of it, I left the conservative cpufreq governor in instead of the performance one. sigh it’s been late …

hmm…

this image is something to ponder :huh: :blink:

I wouldn’t try to draw any conclusions from this. See the discussion above. This benchmark has been commenced in a controlled, yet entirely unfit environment ;)

I don’t think we’ll ever update phpmark. As you’ve mentioned, it takes time and we’d better focus on Yii itself.

I’m wondering if it might be possible to offload the creation and maintenance of sample applications to the framework devs and introduce a github service similar to travis-ci. This would allow automated benchmarks and statistics. Is phpmark.com held by qiang?

Yes, it is.