First Render Time

[Marcus]

Speaking if 40 ms time difference: Did you ping the Servers? This could be the network latency.
You can't meassure - or worse compare - things if your meassurements are less accurate than your results.

I used waterfall diagrams from webpagetest.org. If Ryan would post his debug time from forum start page this would be a great help. There are lots of disadvantages using this, but I don't have access to the forums directly. I have explicitly displayed all my results in my thread and I am happy if someone would provide some data from their community, too.

 

[Marcus]

http://dl.dropbox.com/u/693961/xenforo_first_time_load.mov
17mb .mov (vlc plays it just fine)

I don't have audio at the time, I do not know what you compare and what results you have. Could you help me out there? It is not that fair to compare xenforo.com/community with vbulletin.com/forum as the latter one uses SSL, of course this is slower. I'd like to see both software on the same environment and how this works out. As my install of both vb 4 (live) and xf (test install without visitors) on my server.

 

[qpa]

That doesn't answer my question.
I think your efforts are good and COULD result in finding (and therfore fixing) possbible performance issues,
but you can't just compare "something".
What is the network latency noise to your server (I would suggest to use RMS values)?
What is the vBulletin speed noise?
What is the xenForo speed noise?

And so on.

We are not talking about huge amounts. If you measure small amounts,
you have to think about WHAT you are measuring, not only where the numbers are from.

 

[Marcus]

qpa, I don't know all your specific categorisations for measuring speed. I have displayed all data in my threads and would like if someone would do the same. Why is xf forum start page rendered twice as slow as my vb4 page? For me, twice the time is huge. Is this a server issue? Then how would I correct this? I am posting these data here and I would happy if someone would come with "hard facts", too.

 

[Floris]

So when you view the video, you feel that the loading of the tab, then the url in the location bar, it starts resolving the host to the IP, and gets responses from the web server, the browser starts rendering the html/css/js/images/ etc and the page is loaded and usable. Is actually 'slow' ?

even on ssl, it's fast enough.

 

[Marcus]

So when you view the video, you feel that the loading of the tab, then the url in the location bar, it starts resolving the host to the IP, and gets responses from the web server, the browser starts rendering the html/css/js/images/ etc and the page is loaded and usable. Is actually 'slow' ?

even on ssl, it's fast enough.

My vb4 debug modus shows me half the time xenforo debug modus displays me for the forum start page. 0.1301 seconds (xf debug modus) needs xenforo for the forum start page, and this install is a test install with all my vb-data but without visitors. And then my vb 4 with hundreds of visitors and which is highly modded takes 0.06 seconds (from vb 4 debug modus).

 

[Floris]

Not sure what else I can say.

If you're unhappy with XenForo's performance, and you require the extra 40 ms load time when in debug mode, perhaps it's worth considering to not use XenForo and stick to vBulletin.

Personally, when I load a page in my browser and it loads fast, and doesn't feel sluggish, I am happy.

 

[SneakyDave]

For some reason, I got an alert that this thread topic was updated, but I have never seen it nor watched it before.

EDIT: Brogan explained it

 

[Paul B]

Some posts from another thread (which you did reply to) were merged into it.

 

[Floris]

btw, Marcus, sorry to sound a bit more negative than intended with my post. Simply saying that I don't know what else to say.

 

[jmurrayhead]

Some posts from another thread (which you did reply to) were merged into it.

I was trying to find the suggestion thread, but this is a perfect example of where a notification for threads that have had moderation action taken on them would be great. Example, "The thread, 'You are unworthy', that you have posted in has been merged into the thread, 'Slow first render time'". Would save tons of confusion.

 

[qpa]

qpa, I don't know all your specific categorisations for measuring speed. I have displayed all data in my threads and would like if someone would do the same. Why is xf forum start page rendered twice as slow as my vb4 page? For me, twice the time is huge. Is this a server issue? Then how would I correct this? I am posting these data here and I would happy if someone would come with "hard facts", too.

But that's useless!

You can't compare the numbers like you did. They are too small, to just "look at the numbers". Even the "debug" values you are referring can have significant different values depending on the algorithmic behind.

If it is something like "Ok xenForo needs 2 seconds, but vB4 only needs 1 second" - so your difference is 1 second - it is ok, to just state:
there has to be an issue. 1 second difference is large enough so that other influences are insignificantly small.
But with one-tenth of milliseconds that's just not possible.

Remember that no matter what you are doing: You measure time over a whole IT system.
Possible influences: Network-Infrastructure, Routing, Hardware, Webserver-Software, Web-Applikation.

So first step: Be aware what influences are present.
Second step: With the remaining influences, look how much they (can/could) influence your result.

 

[Marcus]

btw, Marcus, sorry to sound a bit more negative than intended with my post. Simply saying that I don't know what else to say.

I am also curious why this is the case. The first answer I got was that this would be working as designed and this is more a suggestion. But twice as slow as vb 4 - that should not be as Kier told me. So I think I should open up a support ticket.

 

[qpa]

Marcus, regarding the "debug" timing values:
Did you compare the "timing output" algo of xenForo and vBulletin? Are they starting and ending at the same "events?" Are they using the
same clock or counter?

And remember that there are two main kinds of timing: absolut timings and content/load-dependent timings. Sometime one kind of algorithm has a higher absolute timing, but with high loads it performs better.
For example:
1s base + 0.01s per user vs. 0.5s base + 0.1s per user

I'm not saying this is the case here

Perhaps you should "simulate" a bit of load, so that the time values rise, so that you get higher time values.

 

[Deebs]

Marcus,

Can you run your test again and then click the details tab and paste the contents of "Request 1" up until the line that says "Request Headers".

Request 1:

URL: http://blah/
Host: blah
IP: xxx.xxx.xxx.xxx
Location: United Kingdom*
Error/Status Code: 200
Start Offset: 0.001 s
DNS Lookup: 63 ms
Initial Connection: 49 ms
Time to First Byte: 154 ms
Content Download: 106 ms
Bytes In (downloaded): 7.7 KB
Bytes Out (uploaded): 0.3 KB
Request Headers:

 

[Marcus]

Perhaps you should "simulate" a bit of load, so that the time values rise, so that you get higher time values.

Thanks for all your tips! Unfortunately I am not very familiar with website speed. All I use are simple waterfall diagrams and then I use the data the forum software provides me with in their debug mode. It is possible that xenforo and vbulletin 4 measure the time the debug mode displays completely different. I do not know. Only a very advanced xenforo and vbulletin coder could answer this question.

Deebs, thansk very much for your data! This is exactly what I'd like to see Unfortunately I only have one current xf install and to see some better results, I removed 99,9% of my forum data to see if there was a difference in page creation time. The time the page was created from the server was reduced from around 0.14 seconds to a mere 0.06 seconds - which is stilll slower than my vb 4 live configuration for guests but it is definitely ok. I will import all the data from my vb 4 forum to my xf install soon. Here are the values from webpagetest.org for my actual - nearly empty and therefore very fast - xf forum:

 

[Deebs]

Marcus,

I need the specific detail, the time to first byte includes the time to do a DNS lookup. Click on the detail tab and paste the response.

 

[Marcus]

These are the request details:

 

[Deebs]

Is there anyway you can get to the homepage without going thru the initial redirect?

 

[Marcus]

Is there anyway you can get to the homepage without going thru the initial redirect?

Yes but I don't think it matters for these data as you can just cutoff the time it needs for the initial two redirects. (You have to remove the yellow lines, or start by the arrow I painted).