4 ms·
Show HN: PerfAudit – Performance audit of caniuse.com
- roothacker 12y agoLiked the concept of performance auditing websites, are you guys working as consultants?
- chinchang 12y agoRight now we are just looking for awesome websites to audit as case studies and share with developers. Open to audit requests.
- sqs 12y agoWe'd love an audit of https://sourcegraph.com https://sourcegraph.com! :)
- axemclion 12y agoYou could try using http://perfmonkey.com http://perfmonkey.com to run a simple scroll test on your webpage. If you are looking at more powerful tests, try out http://github.com/axemclion/browser-perf http://github.com/axemclion/browser-perf. Here is the result I got - http://www.perfmonkey.com/#/trynow/results/travis/53664194 http://www.perfmonkey.com/#/trynow/results/travis/53664194 Here is a test I ran on bootstrap CSS library over all its versions - http://axemclion.github.io/bootstrap-perf http://axemclion.github.io/bootstrap-perf
- sqs 12y agoCool, thanks! We get a thumbs down for mean frame time. It'd be great to know where we fit (%ile). Do you think it's because we use Bootstrap and SVGs?
- chinchang 12y agoThis is what we are trying to solve. Using automated tool you know that "your site is painting slow or loading slow". But "Why" is something that is still missing in such tools.
- apoorvsaxena 12y agoThanks @sqs, we'll revert back to you soon with #perfaudit of sourcegraph.com
- damienj 12y agoNot as detailled as a manual one, but here is a good start : https://www.dareboost.com/en/report/550828b4e4b074fd5732f006 https://www.dareboost.com/en/report/550828b4e4b074fd5732f006
- strommen 12y agoCool to see someone else doing this. If I was doing this myself, I'd suggest working on js.php. It's taking over 1s to get that single file, which appears [1] cacheable. An nginx reverse proxy or something comparable would make this way faster. If there are uncacheable parts of this file, then those could be separated out. It is interesting that the script is loaded async with "defer", so the page is functional (but not ajax-y) before it's loaded. So the long load time isn't quite so bad. But I'd be interested to hear why the whole file isn't just cached. Also, for a single-page website like this, where all navigations are loaded with AJAX, I'd suggest inlining all the CSS and JS. Finally, I'd think about making the entire page itself cached. There doesn't appear to be any sort of login, so it's probably feasible. But I'm not sure how the Settings stuff is implemented; if that's persisted server-side then caching the page wouldn't work. [1] The Cache-Control header specifies a max-age but not public/private. I honestly don't even know what semantics this has - but it's probably best to make it explicit.
- apoorvsaxena 12y ago@strommen, [1] agree with making static resources cacheable and even using CDN for the same. [2] inlining all CSS and JS, might not be a good idea, as it has several issues related to the same e.g. not being cacheable to start with, parallel resource request etc. There are several other optimizations possible to reduce the page load time that we could share, though the issue in hand which required immediate attention and seemed to be causing a serious damage to the user's experience was its rendering performance, which we have focused in our discussion.
- axemclion 12y agoYou could also try http://perfmonkey.com http://perfmonkey.com for automating these perf audits. If uses the same tools used by Chromium for finding perf telemetry - http://axemclion.github.com/browser-perf http://axemclion.github.com/browser-perf. Here is a run for canIuse using the automated tool - http://www.perfmonkey.com/#/trynow/results/travis/53663113 http://www.perfmonkey.com/#/trynow/results/travis/53663113
- laacz 12y agoWanted to get an example on one of my projects. Sadly it requires write access to my public repositaries. Sorry - no.