I thought establishing a Tunnel was as much for optimizing the network path as it was for security.
I thought establishing a Tunnel was as much for optimizing the network path as it was for security. I guess not.
Pit's separated because it's tremendously useful for zero trust
Pyep, because traffic goes over cloudflare's dark fiber instead of the public internet
IDo you mean unlit fiber?
IAs in, fiber they connect themselves?
EYou're welcome to bench it for yourself, someone in this channel once claimed that using a tunnel did give them a small latency improvement.
Tunnels used to be tightly integrated with Argo (To the point where you couldn't use one without the other) and recently they did work to decouple them into two separate products which is what you see today and why setting up a tunnel doesn't give you the dramatic (but expensive) latency improvements that Argo does anymore
Tunnels used to be tightly integrated with Argo (To the point where you couldn't use one without the other) and recently they did work to decouple them into two separate products which is what you see today and why setting up a tunnel doesn't give you the dramatic (but expensive) latency improvements that Argo does anymore
EArgo is only so effective and expensive because there is a lot of clever routing logic going on in the backend, its constantly learning and adapting to squeeze every last ms saving out of connections
PCF's private backbone network
IYeah.
PThat's why imo it'd make more sense to pay for argo per mbps instead of per bit
K
EArgo will adapt and go whichever routing path is faster at the time
EI think that would be much more confusing
IYeah... I would tend to agree.
EYou'll notice a lot of Cloudflare pricing decisions are done with simplicity in mind, and this is no exception
EWhere did you find that chart? I'm trying to figure out my benchmarking methodology that won't be screwed by delays in the origin app, and not sure how to enable/disable Argo routing for a specific Tunnel/subdomain. I can see only one big On/Off button for the entire account.
KIt’s only visible with Argo on and it is indeed per zone
EAnalytics > Performance and yeah per zone when Argo has been enabled for enough time
KI wish they’d do it on a record level
KMain site? Yeah
Subdomains used by internal departments? Don’t care about speeding those up
Subdomains used by internal departments? Don’t care about speeding those up
ESometimes I wonder if it counts traffic towards the Pages applications as Argo traffic or not
EThe only thing that isnt a tunnel on the domain is two subdomains that serve Pages projects
EOhh, found it. That high level of data makes it impossible to benchmark 
EIts intended for a quick "at a glance" overview rather than concrete data
EIt does let you export it as CSV though which is cool
EWe have AWS Global Accelerator routing the origin (for the same reasons as what Argo should do) and I'd love to benchmark with/without Argo/Global Accelerator to see what works for us.
ENever heard of that product but from a glance it seems to be the AWS version of Argo?
EYes, same concept. Get to their private network ASAP.
EAnd it's a fight I want to see - which backbone routes our users faster, Cloudflare vs AWS edition.
EActivating both probably will tell us nothing - both companies are in all datacenters, so they'll route packets to each other at random
EIf I understand it correctly having both enabled would use Argo to route to AWS's global network which.. wouldnt really achieve anything, and then AWS Global Accelerator would take over and do the actual routing, rendering Argo completely useless
EBenchmarking both would be really cool to see but unfortunately would be difficult to do reliably
EI think so too. So I'd want to measure performance of "Edge -> CF Argo -> CF Tunnel -> Origin" vs "Edge -> AWS Global Accelerator -> origin". Seems this will be difficult to do now if I have only one on/off switch. Will need to create a different account.
EYes, benchmarking the network can be like reading tea leaves. For this case, I'd setup a simple echo page for each network route, and run WebPageTest / SpeedVitals on it.
Kuse spectrum :^))))
Kvery cheap
EI care somewhat. Right now, to my best measure, routing everything through a Cloudflare Worker adds 80-250ms latency (as compared to routing directly to origin). This is acceptable, but 500ms wouldn't be. Our specific use-case is an SPA with API calls, so it's mostly plaintext and relatively little bandwidth.
Kthat's not very orange cloud of you
K
JWait, that’s not kian
Jthat’s Kian with reversed mouth
K
DI envy all these custom(?) stickers & emojis