wp-fan

@acp73 / Anthony,

My pleasure, Anthony.

Are you up to complete perfection? Here is one more small and quick step to do:

Check your site with tools.pingdom.com

In “waterfall” view, check for all entries that start with https://www.anthonypeyper.com/ . These are items in directories which are still not picked up by cloudfront for some reason.

In order to get all pictures and elements delivered by amazon, login to your wordpress backend, go to
performance –> CDN –> scroll down to “advanced”, look for the “custom file list box”.

When I look at https://www.anthonypeyper.com/music/wound-up-tight-au-magazine/ in pingdom tools, for example I see

blog_bg_inner_1358462897.png
is still delivered from
https://www.anthonypeyper.com/wp-content/uploads/p4/images/blog_bg_inner_1358462897.png

To change this, add
wp-content/uploads/*

to the custom file list box.

Do exactly the same with all entries in pingdom waterfall-view that are still delivered via https://www.anthonyheyper.com. Find out in which directories those elements are located and add all those paths to bye custom file list box with a * at the end, just follow the scheme above. You will see that all files will be picked up and delivered by cloud front. This will enhance your site speed even more. (Could be that the line mentioned above catches all those entries already, please double check for yourself).

By the way, pingdom tools gets you very good marks for your site (95/100). I do not know about Yslow grades, I don’t use them. More important to me seems the delivery speed, and pingdom says your pages are being delivered in 1,1 seconds to their Amsterdam service. This is fast. I guess you could go down even a little bit further if you follow the steps described above.

Let me know how this works for you.

Oh, and i personally have decided not to use official version 0.9.2.7. This seems to be rather buggy. I do not think our sites will get much faster than we have got them now, anyway…. :=)))

For all readers of this thread: I think wmiles forked version 0.9.2.5 is the best, fastest, feature richest and most stable version out so far. Check the release notes here and here what the main differences of the official version are.
Non of wmiles enhancements seem to have been introduced in the official version 0.9.2.7. It is a complete mystery to me why not.

@ sLa NGjI’s

Thank you for clarifying.

All of the points raised were already in the roadmap and the contributions here have simply moved them up the queue.

@frederick Townes:

Could you please clarify, Frederick, because I am not sure wether I misunderstood you: Were you able to pick up and include the contributions from wmiles in version 0.9.2.6 or 0.9.2.7?

I have not upgraded myself and will not do so. Comments in this forum hint to severe issues that need to be fixed before I will personally upgrade.

I do agree that the 0.9.2.5 fork from wmiles is the best version released so far. I will not put my website speed at risk with a faulty “official” update.

I have emailed wmiles if he will release an enhanced version based on 0.9.2.6, but have not received any answer yet.

Try this tutorial to set up cnames with your hosting provider, cloudfront and w3c total cache.

I mentioned it is better to setup 10 cnames, right? This is because wordpress aka total cache will request your files from several different cloudfront servers, not only from one. This adds up to speed again – at least this is what I have read on the web and this is my experience, too.

I am based in Germany, your website speed feels better already over here. And I am very confident it will work even faster after you have implemented all steps in this thread. But don’t trust me – trust pingdom :=)

s that forked version different from the official update that fixes the security issue with db caching? if so, this is going to be a maintenance problem.

The author of the modified, repaired and enhanced version, wmiles, says in his release notes that his version is based on the official version 0.9.2.5.
In other words: The security update is included, but he made some enhancements (check the release notes and read the links in his post carefully, you will get the full picture if you read).

I cannot see how this would affect maintenance issues in any way. On the contrary: wmiles version works far better than the official version – for me and other people in this forum.

On the link you gave: https://www.wmiles.com/2013/01/03/w3tc-cache-preload-for-version-0-9-2-5
it states to update the plugin code,

You can download the whole plugin in a modified version from that very page and replace the existing version. Please read carefully.

On the bottom of the same page, there is a detailed step by step guide on how to upgrade to the modified version.

So not sure what to do about “Add Expire Headers”

Check this guide for some good settings.

My recommendation for the “preload cache” settings, once you have installed the modified version linked above:

Garbage Collection Interval: 86400
Cache Preload Update Interval: 300
Pages per Interval: 50

My pleasure, acp73.

It just seem like Cloudfrotn needs S3

No, it does not. in AWS console, you can give your Cloudfront instance a freely chosen name, no need to distribute a S3 bucket. Cloudfront will search and store the files by itself, no need to store them in a S3 bucket. I know many people get this wrong, but believe me, I have the same setup and it works very well.

To summarize things: From my experience the biggest impact has

– precache loading (the original version 0.9.2.5 from the official plugin repository has a fault, use the one that I have linked above instead!)
– using a CDN
– _manual_ minifying (and leaving out the .css and .js files that won’t work with minify)
– using all browser cache options

Kind regards,

Thomas

@acp73:

Thanks wp-fan! Does this mean I have to cancel everything with S3?

Yes. Delete the S3 bucket and use the cloud front upload option in total cache.

Grade F on Add Expires headers

Please use the expiry settings of the screenshots shown in the post that I have linked above. And and recommend to use the w3 total cache version that I have linked above!

You can specify CDN hostnames in your preferences

I have checked your site with tools.pingdom.com . In the “waterfall” tab, I can see some of your pictures are still not being served by cloudfront.

I recommend using 10 CDN hostnames (for example, cdn1, cdn2, cdn3 …). In the beginning, I have only used one, and with 10, more elements should be picked up and served by cloudfront.

In your sourcecode, I can see you have not registered the cloudfront CNAMES with your hosting company. While this does not make any difference in speed to my knowledge, it just does not look nice and may not be beneficial to your SEO. Please check this description how to change this.

Deleted all the css files in the help mode and still the pages break in Minify Manual Mode!

Maybe the problem is with one of the .js files? Try to delete suspicious .js files from minify setup. When you find the culprit, re-add all other .js and .css files, but leave out the .css file which is related to the faulty .js file.

And one more time: Use the w3c total cache version that I have linked in my post above. Definately switch on preload cache with the settings in the screenshot that is part of the post I have linked to. This will make your site a _lot_ faster, too.

Let us know how you are proceeding!

PS: Great photos, by the way.

PPS: Picture missing here. Pitty…

Some suggestions:

– Do not use Amazon S3 but Amazon Cloudfront instead. S3 is not a real CDN, but only a “bucket”.

– Use cache preloading. The official version of “3c total cache has some faults in cache preloading, use this version instead. I use it and it works brilliantly.

– switch minify to manual and use the “help” wizzard to exclude those .js or .css files which “break” your site. Simply leave those files out.

– if you are on shared hosting, do not use database caching, sometimes object caching causes problems.

Using minify is always a bit tricky. I experience on my site that not all .css or .js work after being minified. But, from my experience, you do not have to switch minify off all together.

My solution:
1. In “General” tab, switch minify from “automatic” to “manual”.
2. A help-button will show up in “Minify”-tab.
3. Push “help”-button in minify tab, select the .js and .css manually which can be included in the minified file. Deselect those .css and .js files of those modules which did not seem to work correctly in automatic mode.
4. Check the result on your site.

Make sure logged in WP admins do get to see the minified results instantly (there is an option on the minify page to toggle this behavior).

Kind regards,

Thomas

On Shared hosting, database caching in basic mode is generally not recommended, as far as I know, because performance issues seem to be frequent in this case.

@strats100 and WPJEDI:

The forked version of wmiles mentioned and linked in my post earlier changes the cache behaviour all together.

It checks the sitemap.xml for new posts and checks wether these files are in the cache or have changed recently. If they have changed recently or have been removed from cache, it will immediately re-cache these pages. It will also rewrite the files in cache before they expire. And garbage collection actually works as it should.

To put it in a nutshell: With this setup, there are pre-cached files present reliably _at all times_ .

My settings:
Gargabe collection interval: 86400 seconds (every 24h)
Cache Preload Update Interval: 300 seconds (every 5min)
Pages per interval: 50 (I am on decent shared hosting, this value works very well in my case).

This setup made my website react 100% faster. With minify enabled and set up correctly and using Amazon Cloudfront as CDN, my site now loads in 1,2 seconds on first call in average (Europe), 2,2 seconds in the US (measured by tools.pingdom.com)

Regards,

Thomas

for me its unfeasible to switch to another fork due to maintenance considerations.

I understand.

Just for your information: When it comes to update notifications, the fork mentioned above behaves like a “normal” version of 0.9.2.5 . You will be notified instantly if a new “official” unforket version is being released.

The fork you mention seems to have been developed for a different problem.

@yousefed: Yes and no. It has been developed to fix other problems, but addresses the garbage collection challenge, too. In my installation, I use preCache-Loading, and I can confirm that garbage collection works 100% .

But of course, people will be interested, if you find a fix for the official version and share it with us…