First Numbers from CloudFlare



  • It's been one month since CloudFlare has been in front of MangoLassi and in the interest of being transparent and this just being fun and interesting, we are going to grab the analytics and show them here.

    We could have waited a few days to get better numbers because the first two days of CF were "ramp up" as we started getting more traffic. But why wait and hide that, this is great at showing how things have improved from what little we know from before this time. These numbers include literally the first full day of analytics from CloudFlare so nothing is missing (the day before this was a half day where CF was turned on partway through the day.)

    0_1465454033775_Screenshot from 2016-06-09 09:28:48.png

    That's 11.778 total requests through CF! As you can see, caching doesn't do a lot for us, but it does something for sure. You can also see how after the first two days, our sustained traffic levels are higher so the first two days, and maybe part of day three, were still the system getting up to speed. Also the cache warming up.



  • 0_1465454127646_Screenshot from 2016-06-09 09:29:09.png

    Here is the traffic graph. The bandwidth numbers are quite different and show a large increase in total content being pulled by roughly the same number of requests. Very interesting. The cache remains pretty flat while the total really climbs.



  • 0_1465454202102_Screenshot from 2016-06-09 09:29:26.png

    Unique visitors is great. Nearly 30,000 unique visitors for the month! This one has a very visible weekend / weekday cycle. And again, shows growth towards the end of the month. Awesome!



  • 0_1465454297323_Screenshot from 2016-06-09 09:30:19.png

    US is the big traffic source, no surprise there. This map makes it hard to see where traffic comes from, but there it is.



  • 0_1465454329197_Screenshot from 2016-06-09 09:30:34.png

    Sadly, this is all of the details that we get about per-country traffic. China is the source of nearly all of the bot traffic so that explains why China looks so huge. Serbia gets talked about a lot, perhaps that has caused us to show up there enough to get a lot of traffic? That Australia beats the UK is really surprising.


  • Service Provider

    CloudFlare's numbers always seem huge compared to every other analytics tool I've seen. Do they just over-inflate the values or not count them right?


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    CloudFlare's numbers always seem huge compared to every other analytics tool I've seen. Do they just over-inflate the values or not count them right?

    Or do they show realistically what's there? CloudFlare showed us gaining numbers rapidly once it was in front and HTTPS was turned on. It shows requests, the other numbers that we have are heavily filtered (we can watch them go down every hour as certain types of traffic are removed.) The CF numbers are the only ones that show raw what is happening including cache hits and such that don't make it through to the application itself.


  • Service Provider

    From what we can tell, and it is a pretty huge guess, using CF caused our traffic to increase by 50-100%. It makes the site faster, improves SEO and it is when HTTPS was enabled so it's not at all unlikely that it had a huge effect.


  • Service Provider

    But which is the more honest appraisal of traffic. What is the right way to monitor engagement levels on the site? Why do the other providers have harsher filters with lower numbers?

    The math makes me nervous, you'll naturally want to believe the higher traffic numbers, my own website shows really high CloudFlare values but all the other tools are more discouraging.

    I'm not convinced that CloudFlare gives you any boost in traffic or SEO. What else are you using for analytics? Has that equally gone up?


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    I'm not convinced that CloudFlare gives you any boost in traffic or SEO. What else are you using for analytics? Has that equally gone up?

    CF would have to be lying for that not to be the case. The numbers increased over time as CF was applied. For four days there was solid increases, which is about what we would expect from a change that impacted SE rankings. We definitely got an increase in reports of being high on SE results after moving to CF.

    Other analytics have not gone up, but should not, because they should have gone done if the traffic was flat because some of the traffic was offloaded to CF. CF absorbed nearly half a million requests during the month that didn't come through at all. That's pretty huge.


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    But which is the more honest appraisal of traffic. What is the right way to monitor engagement levels on the site? Why do the other providers have harsher filters with lower numbers?

    None are everything on their own. There is no right way because nothing can tell you engagement. Nothing measures the amount of time spent actually reading, pondering, scrolling, etc. So the best you can do is use a combination of tools.

    CF is good because it is standard and equitable. Even if it is on the generous side, the collection is uniform.

    NodeBB errs on the farthest other side, the most conservative possible numbers. It does things like ignore time on a page, throws out refreshes and such. So a lot of behaviour that every other site would consider a page view, NodeBB collapses into a single event. It's the most conservative numbers that I've ever seen and the only platform that I've ever seen "clean up" to reduce the numbers once it has data on where the traffic came from.

    So the one is clearly liberal, it shows all traffic. The other is very conservative, only showing traffic it feels meets certain criteria.

    Putting the two together gives you the best picture, but even that is full of guess work. What do the numbers represent? Are people popping in and leaving? Are they hanging around and reading stuff? Are they reading one post and moving on? Are they scrolling down and reading post after post? We just can't determine that because those things don't include "clicks".


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    Why do the other providers have harsher filters with lower numbers?

    Most have little to no means to see all of the data. Google Analytics, for example, has a lot of data but misses a lot of the traffic. It can't read what is going on inside of the NodeBB platform. So it misses things like moving from one thread to another. Anything on the server misses anything picked up the cache and so forth. CF is unique in collecting the data first, before different systems pick up the parts.


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    What else are you using for analytics? Has that equally gone up?

    What else is there? Not sure what else is capable of collecting the data.

    I should say... if early June is an indicator, this will be a record month. But it is far too early to tell. But it is possible to know what causes increases or decreases in traffic month to month. It fluctuates a lot so very hard to determine individual causes.


  • Service Provider

    @scottalanmiller said

    So it misses things like moving from one thread to another. Anything on the server misses anything picked up the cache and so forth. CF is unique in collecting the data first, before different systems pick up the parts.

    Doesn't google analytics track by the URL as nodebb is not 1 single page. From what I've seen of it you can monitor real data on your users. Gender, age groups, other interests, bounce-rate, how long they spend on each page, what's the record page where viewers remain.

    If I say my website has X number of visitors per day based on an extremely broad filter on what is considered a unique visitor. Should I feel confident in telling the world those stats?

    @scottalanmiller said

    What else is there? Not sure what else is capable of collecting the data.

    Well, what does Google Analytics think? What are the numbers from that? Can we weigh the conservative V the liberal stats?

    There are a huge number of other analytics tools as well, some of them are incredibly invasive, they can even give you the name of the business who looked at your website based on the external IP, show the pages you read, show who is who at the company, lots of data.


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    Doesn't google analytics track by the URL as nodebb is not 1 single page.

    It is one single page, actually. The URL changing is part of the REST API, but does not denote page loads. So GA can't tell what is happening.


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    If I say my website has X number of visitors per day based on an extremely broad filter on what is considered a unique visitor. Should I feel confident in telling the world those stats?

    CF and NodeBB have very close numbers as to unique users, as they should. So that looks to be very accurate.


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    Doesn't google analytics track by the URL as nodebb is not 1 single page. From what I've seen of it you can monitor real data on your users. Gender, age groups, other interests, bounce-rate, how long they spend on each page, what's the record page where viewers remain.

    It collects user data, yes. But information about their engagement it cannot gather from a custom web application that has no normal web server.


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    There are a huge number of other analytics tools as well, some of them are incredibly invasive, they can even give you the name of the business who looked at your website based on the external IP, show the pages you read, show who is who at the company, lots of data.

    Yes, but can any of them give data on the traffic? Google can't. I don't know how anything could. NodeBB itself can gather one type of data and the CDN (CF in this case) can gather another. I'm not sure how any tool could give more accuracy without tapping directly into one or the other of those.


  • Service Provider

    Well what traffic data are you looking for?


  • Service Provider

    @Breffni-Potter said in First Numbers from CloudFlare:

    Well what traffic data are you looking for?

    All of it would be nice :) I'm sure Google has some interesting data, even if the volumes are incorrect. The stuff we are discussing here though is just volume, and CF and NodeBB likely have the best numbers. Most importantly, they have the only sources for accurate ones.



  • There we go.... a full 12m!!

    0_1465540674585_Screenshot from 2016-06-10 09:35:58.png



  • @mlnews said in First Numbers from CloudFlare:

    0_1465454297323_Screenshot from 2016-06-09 09:30:19.png

    US is the big traffic source, no surprise there. This map makes it hard to see where traffic comes from, but there it is.

    So, what happened to Brisbane, Sydney and the entire nations of New Zealand, the Solomon Islands, Vanuatu etc ?
    ...
    (Just being difficult) ;)



  • @mlnews said in First Numbers from CloudFlare:

    Serbia gets talked about a lot, perhaps that has caused us to show up there enough to get a lot of traffic?

    I'm pretty sure there is one person for the cause of that. :-)



  • @mlnews Hi; I work for CloudFlare. Just wanted to share a few tips that might be able to help your performance:

    1. HTTPS - By enabling HTTPS you can allow your users to access your site through HTTP/2 and SPDY, which dramatically improves page load speed. After this is enabled, you can configure HTTP/2 Server Push to get things to be even faster.
    2. Page Rules - Page Rules allow you to use a "Cache Everything" setting where you can even cache static content.
    3. On your server be sure to whitelist CloudFlare's IPs to make sure they're not rate limiting your connection to us.
    4. Ensure correct cache headers are being sent to ensure the content you want to cache, is being cached by us.

    Hope your enjoying our service so far! Let us know if you need any further help.



  • @IcyApril said:

    1. HTTPS - By enabling HTTPS you can allow your users to access your site through HTTP/2 and SPDY, which dramatically improves page load speed. After this is enabled, you can configure HTTP/2 Server Push to get things to be even faster.

    lol. How long has this been on @scottalanmiller todo list? It took us 2 years to get email, hopefully we can get this done faster?



  • @aaronstuder said in First Numbers from CloudFlare:

    @IcyApril said:

    1. HTTPS - By enabling HTTPS you can allow your users to access your site through HTTP/2 and SPDY, which dramatically improves page load speed. After this is enabled, you can configure HTTP/2 Server Push to get things to be even faster.

    lol. How long has this been on @scottalanmiller todo list? It took us 2 years to get email, hopefully we can get this done faster?

    We already have HTTPS here at ML, you just have to do it on purpose, it's not forced upon you.



  • But when I type HTTPS I get this.

    0_1465561358860_ml.png



  • @Dashrender I am not happy until it is enabled by default, and I get a green lock :)



  • @Dashrender said in First Numbers from CloudFlare:

    But when I type HTTPS I get this.

    0_1465561358860_ml.png

    Probably because the images are hosted by imgur?



  • @coliver That's my thought :(


Log in to reply
 

Looks like your connection to MangoLassi was lost, please wait while we try to reconnect.