Testing Argo Performance

argo

#1

I added Argo to a high volume client site today. It’s supposed to save 318 minutes of latency per month, as per the emailed message. The site is pretty fast already though so I’m skeptical. 19,080,000 milliseconds saved across 52,561 page loads (creating 2,163,163 requests) per month = 363ms saved per page load.

GA says the average page load time is 4.6s. If all goes well, that should decrease to about 4.2s.
I look forward to the results! And I’ll share them here next Friday (if I remember).

If I may be a cheap jerk for a moment… why is Argo something we have to pay extra for? The price of Pro was tripled not too long ago. And Argo seems like a maintenance upgrade to what a CDN is supposed to be - optimized routing of traffic. Some of this stuff is above my head so I could be missing something elementary!

09 PM


Introducing Argo — A faster, more reliable, more secure Internet for everyone
#2

I’ve got it enabled on a big site also, and I can’t see any noticeable difference in google analytics (if anything page speed is lower)

How does Cloudflare suggest we check if there is an improvement or not?


#3

this is where extended analytics would come in handy DNS Analytics extended to Page Speed Analytics? :smile:


#4

Today when I want to check I use catchpoint. If you’re an enterprise customer you might ask your account team if they can run a test for you. For other folks… catchpoint is a good way to test, a better way to report from Cloudflare is being worked on but presenting the data in a way that is meaningful is more challenging than one might think.


#5

@jules I moved this to its own topic so people can find it easier.


#6

I also just realized I never addressed your pricing questions. Sorry about that. Didn’t think you were being a jerk…it’s a common question.

Not exactly. Pro price hasn’t changed, but discounts on additional Pro zones on the same account were removed.

Similar but also, not exactly. :slight_smile: Matthew addressed it in a variety of tweets, and here are some good quotes:

“we end up paying for ingress/egress multiple times for the same packet as it traverses the Argo virtual backbone.”

and

“traditional CDN just speeds up static content, Argo speeds up dynamic content that must be fetched from origin.”


#7

Yes that makes sense, thank you. Since making my post I’ve read up about Argo and how it works. It’s not just a build cost CF absorbed, but also has to pay for it’s traffic ongoing.

I remember Pro was $5/month/domain when I started using CF. Now it’s $20. Am I mistaken?


#8

The price for a Pro zone has always been $20. There was a discount for additional Pro zones on the same account, bringing additional zones down to $5 each. It was phased out about a year ago, with all existing discounted zones grandfathered in at the discount price.


#9

I was getting the green check mark for CDN on webpagetest until I installed argo, and now I’ve got the black X; not good. What’s up?


#10

It’s been over a month. Here are the comparison stats.

In GSC I see only small speed gains. But the graph is still showing non-argo data and there’s no way to filter that.

46 AM

The only thing that improved in GA is the domain lookup time. Everything else is slower. :confused:

Maybe I had the site so well optimized otherwise, that there wasn’t anything for argo to improve?
Or since the server is centrally located in the US and all traffic is US, argo doesn’t help much?


#11

if all traffic is US then probably… tried breaking down speed stats in GA by country to see if helps outside US ?


#12

My performance with Argo is the same: it doesn’t seem to do what it should do (make the website faster). My redirection time did improve, while the page download time got much worse. Not sure what to make from these numbers.

My origin server is in Germany while most of my traffic comes from the US. According to Cloudflare Support, Argo performs better if there’s a longer route between the Cloudflare data center and the origin server.

I think the distance US -> Germany should be long enough for Argo to optimise the network path. Otherwise it probably defeats the purpose of this new product.


#13

Individual mileage can vary. Not all sites are optimal cases for Argo. That being said, we’d like to take a closer look. Can you file a support ticket so we can get a look at all the settings and analytics? Once you have filed it let me know the ticket number.


#14

Thanks Ryan for replying but I already made a support ticket. I closed that ticket since it didn’t really go anywhere. The conclusion from support was that:

“you have almost all your website files aggressively cached, [so] there might not be a lot of request being sent to your origin to give you any meaningful stats.”

and

“For requests that are routed to your origin, it seems that they are optimized by Argo.”

My take on this is that I don’t get enough uncached requests (which currently stands at 62k per month) to let Argo (and the tiered cached feature of Argo) have a measurable impact on my site speed.

Which is not a problem, because I understand that’s how it works with a network as big as Cloudflare’s. Argo will probably help once my website gets more popular.


Anyhow, I was hoping that Argo with its tiered cache feature would behave something like an origin shield (that I mentioned earlier here) so that it would also help the blogs with modest traffic per month. But it doesn’t do that in my case, yet.


#15

I can have someone review your ticket again, but based on what you described it sounds like the same as what an Argo engineer just explained to me. At this time you are probably not the optimal use case.