Only cache misses with pagespeed

v3
#21

I am talking about the initial TTFB not the overall load time.

My understanding is that http/2 helps with pushing multiple files over a single connection after the TTFB is complete along with the initial handshake, HTTP/2 does not affect TTFB as far as I can tell.

It may have to do with the strong ciphers I am using (official recommended from https://cipherli.st/ to get A+ security rating) or that I am using haproxy with CloudFlare full SSL so encryption is taking place twice (once between the user and CF then between CF and the VPS).

When I use just flexible ssl from CloudFlare the TTFB is much lower presumably because my VPS doesn’t have to encrypt anything.

I will try with nginx doing the SSL with the same strong ciphers to see if that reduces TTFB.

@rsmith4321 and @janiosarmento would you mind sharing your URLs so I can examine TTFB, more data would help :slight_smile:

#22

I think you just answered your question it’s Cloudflare. Think of all the routers between you and your server. I’m fanatical about page load times and http2 adds no latency I can find.

#23

And check my site out https://www.ryansmithphotography.com. I have a very image heavy wordpress site, the entire first page load time is 465ms according to pingdom tools New York server, 63ms ttfb. A+ ssl labs test. It’s not http2 causing your delay.

#24

Thanks Ryan, just to reiterate I am not using HTTP/2 and HTTP/2 should have nothing to do with TTFB. I am talking about the initial TLS handshake delaying TTFB. As I said in my findings SSL TTFB is sensitive to geographical distance (which http is not in my tests). I am not disputing HTTP/2 is a good thing :). I am arguing that the HTTPS TLS handshake does create an overhead for TTFB which is not good for users with an international audience.

I’m guessing your VPS is in the USA somewhere. Compare your TTFB, it is 3x higher outside the USA. We can of course all get good TTFB when the server is geographically close :wink:

New York: http://tools.pingdom.com/fpt/#!/rEK04/https://www.ryansmithphotography.com/ Sweden: http://tools.pingdom.com/fpt/#!/cHwKQc/https://www.ryansmithphotography.com/

Now look at the data here for a site without https where the VPS is located in Amsterdam. The TTFB is less affected (we must expect some latency) Sweden: http://tools.pingdom.com/fpt/#!/eb8dOr/www.htpcguides.com New York: http://tools.pingdom.com/fpt/#!/erZS4/www.htpcguides.com

Of course I need to run more tests but this is the overall trend I have found with SSL TTFB overhead along with other users.

#25