656418 - Mark Up Latency Concerns

Status: VERIFIED FIXED

(Websites Graveyard :: markup.mozilla.org, defect)

Description

[Ryan Snyder [:ryansnyder] [:rsnyder] [:rysny]]

There are some concerns about latency with the Markup site in production.

When viewed from a computer that has yet to visit the Markup site, there is a 10-12 second amount of time that it takes to load the page and have a functional site that can be viewed.

This has not been identified as a backend or frontend issue, it's being filed as an area of concern.

Comment 1

[Stephen Donner [:stephend] Not actively reading bugmail]

It's just one test so far (though I can help run more, from different vendors, tomorrow), but http://www.webpagetest.org/result/110512_AC_KBXY/ shows us doing 10.881s from Dulles, VA, using Chrome 11 over DSL.

Comment 2

[Austin King [:ozten]]

Attachment: Gomez analysis from LA data center

Comment 3

[Austin King [:ozten]]

Attachment: Gomez analysis from Buenos Aires data center

Comment 4

[Austin King [:ozten]]

Attachment: Gomez analysis from Frankfurt data center

Comment 5

[Austin King [:ozten]]

Attachment: Chrome's network tab from Seattle

Comment 6

[Stephen Donner [:stephend] Not actively reading bugmail]

http://tools.pingdom.com/fpt/?url=http://markup.mozilla.org/&id=4629282, too (load time of 13 seconds).

Comment 7

[Austin King [:ozten]]

I'm seeing a wide variance in page requests.
http://pastebin.mozilla.org/1223775

Two classes: 
Slow is about 4 seconds
Fast is about 400 millis

I see the Date and expires headers differ by 30 minutes on the two requests.
Could there be a mis-configured load balancer/Zeues/Apache scenario?

Any other ideas based on the pastebin? The Gomez LA screenshot kinda looks like one of these slow requests, 3+ seconds waiting for the first byte of the page to return.

Comment 8

[Austin King [:ozten]]

(In reply to comment #6)
Cool, be sure to test 'https://markup.mozilla.org/en-US/#/' to get apples to apples.

Comment 9

[Austin King [:ozten]]

Attachment: pingdom fast

Comment 10

[Austin King [:ozten]]

Attachment: pingdom slow

I made towo tests one after another. Not the time scale on fast (7 seconds) vs slow (18 seconds).

Comment 11

[Corey Shields [:cshields]]

Fixed.

11:24 < ozten> cshields: made 30 requests and didn't hit any slower than 800millis

We are now stripping the 'no-cache' header requests that were causing the zeus cache to clear on every request (this defeating the purpose of the zeus cache).

Similarly, I noticed this same problem with mozillademos and it has sped up for me as well.

Comment 12

[Stephen Donner [:stephend] Not actively reading bugmail]

Attachment: Post-fix screenshot

This is only one of several services I've used, but I've verified we're ~3 seconds for initial suck-down, and up to an additional 1 second or so for "complete" rendering.

Comment 13

[Stephen Donner [:stephend] Not actively reading bugmail]

Attachment: Post-fix screenshot, Browsermob.com test results

Comment 14

[Stephen Donner [:stephend] Not actively reading bugmail]

Verified FIXED:

http://www.webpagetest.org/result/110512_N7_KH32/1/details/
http://browsermob.com/free-website-performance-test/550655713ef647e19fb8cd02153a5d38