Home > The Newbie Zone > Questions and Answers

Troubleshooting lander page load times (21)


06-23-2017 05:02 PM #1 plamen (Member)
Troubleshooting lander page load times

Good day everyone,

Oh the joys of affiliate marketing. Stayed up until 5am editing, changing up, and removing redirects on a lander only to find it loads in 3-4secs

I've tried a few things to cut the load speeds, I linked to the google jquery.min.js and I also tried creating a separate file for it. Compressed images. But still slow, by the way this is one of those FB looking spinner sweeps landers. Is it possible to get this sub 1 sec?

Click image for larger version. 

Name:	Capture27.png 
Views:	45 
Size:	43.8 KB 
ID:	15779

This domain is on NS1 and s3/Cloudfront.


06-23-2017 09:20 PM #2 vortex (Senior Moderator)

Which tool are you using to test speed? Pingdom?

Try this tool - it can provide good suggestions on optimizing for speed:

https://gtmetrix.com/

Also - go to the following post and look for the section "Try decreasing your page load speed!" - you may find some good tips there:

https://stmforum.com/forum/showthrea...-AM-FAQ-Thread!

Hope that helps!



Amy


06-23-2017 10:32 PM #3 platinum (Veteran Member)
Troubleshooting lander page load times

In case you have ripped your lander from a spy tool like Adplexity, preliminary optimizations have been already done to that lander. Results on image compression will show you with very little difference when passing the images on them.

When it comes to jquery files keep in mind to always replace them with google respective versions.

On testing tools one thing you will notice on speed timings is that on the first trials everything seems pretty slow. The case may be that the files are not fully cached in the cloudfront distribution server for the desired geo.


Still not happy with timings? Try Google's online storage, maybe it can work better on specific geos.

Also keep a close eye on traffic source and tracker stats. High difference in visits delivered from TS to tracker may indicate that there is something wrong.
Last suggestion, try a different DNS provider, maybe name serving services can be faster, but always keep an eye on pricing and overall improvement.


06-25-2017 02:26 AM #4 plamen (Member)

Yes, that initial screenshot was from Pingdom. The lander's for an offer in AU so that's why I used it.

I since moved this domain from NS1 to Route 53 and tested it with GTMetrix as well:

Click image for larger version. 

Name:	GTMetrix spinner lander.PNG 
Views:	29 
Size:	156.9 KB 
ID:	15787


Top 2 recommendations are GZIP and browser caching. I remember reading somewhere here on the forum that GZIP doesn't make much of an improvement so that's why I ignored that one. And caching wouldn't make a difference for 1st page visit.

To platinum's point, these being ripped from Adplexity, I would think they're pretty well optimized already? Even with that I cut out one JS script which I deemed unnecessary and as far as the minified JS I split tested with one linking to the code (from https://code.jquery.com/jquery-2.1.4.min.js) and one "locally" stored on the CDN. Both versions are 2+ secs on Pingdom in AU. And 1.2-1.4secs on GTMetrix (after repeated tests).

Is it realistic for one of these spinner-type landers to be sub 1 sec with a CDN like Cloudfront?


06-25-2017 06:59 PM #5 vortex (Senior Moderator)

Perhaps upload the lander somewhere so other members can have a crack at it?

I can't promise that will happen, but might be worth a shot!

(Maybe make it into a challenge of some sort - start another thread - and offer a prize?)



Amy


06-26-2017 09:58 AM #6 caurmen (Administrator)

I'd run the pingdom test a dozen times in quick succession and see what the speeds are at the end of that test.

53kb in 2 seconds implies that something in the hosting chain is slow as hell, and if that was the first time you'd run the test, or one of the first few, what was probably happening was that Cloudfront was requesting it from source, and thus being very slow.

Spam it a bit to give Cloudfront the idea it's popular, and see what the load times drop to.


06-28-2017 12:01 AM #7 plamen (Member)

Okay I believe I got to the bottom of the speed problem! It must have been the DNS. After switching DNS over to Route 53 on all domains now and making sure to add the dot (.) at the end of the nameservers in the registrar (had neglected to do this - not sure if it caused delays), the load times were cut in half. I can finally see interactive sweeps landers with JS load around 200-500ms. And that's only after 9 hours of propagation.

I'd like to clarify 2 things on Cloudfront/Route 53 -


06-28-2017 04:11 PM #8 caurmen (Administrator)

Interesting results! Thanks for following up and letting us know what worked.

Incorrect DNS setup can absolutely lead to delays, as more queries are needed than should be.


06-28-2017 08:28 PM #9 vortex (Senior Moderator)

Wow thanks for reporting back! I'm going to link to your post from the big FAQ, and direct other people that experience the same problem, to it.

Congratulations!!


Amy


07-01-2017 01:22 AM #10 plamen (Member)

I believe I spoke too soon!

I followed the Pingdom real time tracking guide and decided to implement it in this next campaign.

Here's what I'm getting from Pingdom real time:

Click image for larger version. 

Name:	Pingdom speed avg.PNG 
Views:	13 
Size:	16.8 KB 
ID:	15839
Click image for larger version. 

Name:	Pingdom real time speed test.PNG 
Views:	22 
Size:	65.6 KB 
ID:	15840
Click image for larger version. 

Name:	Pingdom speed by page.PNG 
Views:	15 
Size:	31.2 KB 
ID:	15841
Click image for larger version. 

Name:	Pingdom speed overview.PNG 
Views:	14 
Size:	48.6 KB 
ID:	15842

Just like with the last campaigns - conversions are looking horrendous right now.


07-01-2017 01:57 AM #11 plamen (Member)

If anyone can share any advice on how this loading speed issue I'd be eternally grateful - this one issue is holding me back from running my first real campaign, optimizing, etc.

I've followed the S3/Route 53/Cloudfront setup tutorials here on STM to a t. All of these landers are from Adplexity, although I've played around with them quite a bit, mostly cosmetically. I meticulously went through them and removed sneaky redirects and replaced jquery files. I'm also split testing with Google hosted jquery files but both versions are still slow.

No idea what is going on. I can only guess that it's coming from DNS (Route 53) issues since at times testing on Pingdom I frequently have the timer run to 60s and waterfall will show a really long DNS time (in pink):

Click image for larger version. 

Name:	fbspinnervibrantlocaljs-loadtime.PNG 
Views:	17 
Size:	32.1 KB 
ID:	15843

And on GT Metrix I will see long "Blocking" times (in brown):

Click image for larger version. 

Name:	GTMetrix waterfall.PNG 
Views:	15 
Size:	61.6 KB 
ID:	15844

Any thoughts on what could be causing these issues?


07-01-2017 01:48 PM #12 plamen (Member)

Day 2 of monitoring with Pingdom and it's gotten even worse. First day stats were actually median, not average, so slightly skewed. I've seen some page loads today go as high as 300+ seconds. The average is now close to 8 seconds. All 4 landers (with Google JS and JS file versions) are performing bad. There's no one that's impacting the average. I will try to change the lander links to go through the cloudfront URL but from what simple speed test tools like Pingdom/GTMetrix are telling me, that isn't the issue.

Click image for larger version. 

Name:	Pingdom speed overview 3.PNG 
Views:	15 
Size:	35.1 KB 
ID:	15852
Click image for larger version. 

Name:	Pingdom day 2 overview.PNG 
Views:	14 
Size:	31.4 KB 
ID:	15853
Click image for larger version. 

Name:	Pingdom speed by page 3.PNG 
Views:	15 
Size:	43.6 KB 
ID:	15854
Click image for larger version. 

Name:	Pingdom browser avg.PNG 
Views:	13 
Size:	50.4 KB 
ID:	15855
Click image for larger version. 

Name:	Pingdom speed avg 3.PNG 
Views:	12 
Size:	17.9 KB 
ID:	15856


07-01-2017 03:02 PM #13 platinum (Veteran Member)

@plamen can you test one of your landers with CloudFront's domain in order to make sure the DNS is not interfering in the performance? Also, S3 bucket location and where it has to be served.
Usually cloudfront requires around an hour to distribute the content


07-01-2017 08:47 PM #14 plamen (Member)

Quote Originally Posted by platinum View Post
@plamen can you test one of your landers with CloudFront's domain in order to make sure the DNS is not interfering in the performance? Also, S3 bucket location and where it has to be served.
Usually cloudfront requires around an hour to distribute the content
Yeah that's exactly my thought, so this morning I edited all the campaign links in Voluum to point to cloudfront instead of the domain name just to see what would happen. Conversions did not pick up after 6 hours like that. I signed up for the Starter package on Pingdom so I can only monitor one site and couldn't get RUM data on the cloudfront url (if that's even possible).

Anyhow, I tested the landers on Pingdom and GT Metrix and see no difference in test times between using the domain name or cloudfront url.

I just tested some of the landers on Pingdom's free tool and some of them showed 100ms-200ms speeds from Australia and some 3-4 seconds (after multiple retests). On GT Metrix they range from 0.5 secs - 1.5 just now.

So I was curious to see how realistic these figures are, especially the lander that shows 100ms-200ms (albeit in Australia). But from today's real-time monitoring, the median on that "quick" lander is 2.6 secs and the average is 7.0 secs. There's some US and Desktop views skewing these so I'm expecting my numbers in the GEO are higher.

Real-time view from just now:
Click image for larger version. 

Name:	Pingdom real time speeds 2.PNG 
Views:	12 
Size:	74.0 KB 
ID:	15857

Anyone with S3/Cloudfront having similar issues?


07-01-2017 08:55 PM #15 platinum (Veteran Member)

One other option is going with a Linode VPS server for your landers, also Google Cloud Storage another alternative.
Split test the platforms and stick with the one that works out for you.


07-01-2017 09:33 PM #16 plamen (Member)

Quote Originally Posted by platinum View Post
One other option is going with a Linode VPS server for your landers, also Google Cloud Storage another alternative.
Split test the platforms and stick with the one that works out for you.
Thanks, I thought a CDN would beat a VPS in most cases. Have you worked with Linode? What are the speeds like if you're not running in one of the GEO's they're data centers are in? For example, from your experience, do you think the Singapore DC would outperform Cloudfront in Indonesia?

If you don't mind sharing, what's a realistic load time to expect with them considering I invest in a 4gb or 8gb VPS?


07-01-2017 10:04 PM #17 platinum (Veteran Member)

ATM can't share any results since i'm on mobile, but one thing i would do in this situation is split teat the three options.

Zeno has already shared a stack script in the below thread that will help you setting up your Linode server if you're not quite sure how to set it up:
https://stmforum.com/forum/showthrea...he-StackScript

Follow the instructions on this guide to have it done in just a few minutes.

The good thing about Linode servers is that you can turn the down and up based on your needs and pricing is on hourly base.
On top of that you can try cloudflare to improve speeds if needed (not quite sure how that might work)

One thing about Amazon S3 that might cause the issue is the storage type selection, in case you've selected near-line/midline storage that might not be the best option. You need to select the high performance one. Check them again maybe we can figure out what might be the issue.


07-01-2017 11:29 PM #18 plamen (Member)

Quote Originally Posted by platinum View Post
ATM can't share any results since i'm on mobile, but one thing i would do in this situation is split teat the three options.

Zeno has already shared a stack script in the below thread that will help you setting up your Linode server if you're not quite sure how to set it up:
https://stmforum.com/forum/showthrea...he-StackScript

Follow the instructions on this guide to have it done in just a few minutes.

The good thing about Linode servers is that you can turn the down and up based on your needs and pricing is on hourly base.
On top of that you can try cloudflare to improve speeds if needed (not quite sure how that might work)

One thing about Amazon S3 that might cause the issue is the storage type selection, in case you've selected near-line/midline storage that might not be the best option. You need to select the high performance one. Check them again maybe we can figure out what might be the issue.
I just got off the phone with Amazon, person I spoke to didn't know of a high-performance option for S3. Only thing he knew about was "Transfer Acceleration" which is for speeding up uploads. I also poked around my account and AWS articles but haven't found anything that would indicate an upgrade option.


07-02-2017 12:11 AM #19 platinum (Veteran Member)

Quote Originally Posted by plamen View Post
I just got off the phone with Amazon, person I spoke to didn't know of a high-performance option for S3. Only thing he knew about was "Transfer Acceleration" which is for speeding up uploads. I also poked around my account and AWS articles but haven't found anything that would indicate an upgrade option.
Here you are: Enter your bucket - Select all files - More - Change Storage Class
Click image for larger version. 

Name:	JVCkzss7TpWUUzCn2aBTqw.png 
Views:	19 
Size:	62.8 KB 
ID:	15858


07-02-2017 01:41 AM #20 plamen (Member)

Interesting - so you're saying change it to "Standard - IA" instead of "Standard" correct?

What are your thoughts on Cross-Region Replication? I'm thinking about trying it out.


07-02-2017 07:22 PM #21 vortex (Senior Moderator)

Wished I was better at technical stuff to be able to help! Hope you guys can get to the bottom of this.

What I WILL do is bump this thread from time to time to hopefully draw more attention and advice.



Amy


Home > The Newbie Zone > Questions and Answers