Would be nice indeed, but that would reveal their data is flawed. In fact, I'd like to see the data per ASN and per subnet. Because you can have an ASN that spans over a huge area (e.g. all of US), where particular dense populations will show everything is fine, but more rural areas will "suffer".
In fact I used to work with a huge adult site, where there were performance metrics down to a /24 subnet level - so in cases like that, performance problems could be investigated quite easily.
That's true. Other than ASN & subnet, would like to see metrics per Cloudflare plans. Enterprise plan have more servers & locations, while the free plan goes to the "cheapest locations"
So the "All networks" TTLB metric for example says CloudFlare is #1 in 5000+ networks, other CDNs are faster in almost 15.000 other networks, meaning CloudFlare is only the fastest in roughly 30% of the networks.
Then the blog post ends by saying CloudFlare is not the fastest in 10% of cases which seems simply false.
The metric selection as well seems to be selected very much in a way where it's skewing the actual results into making them be much better than they are. A much better result would be to actually show the actual global median response times, which would likely be extremely close to each other between all major providers. Meanwhile this post attempts to make CloudFlare look multiple X better than Akamai for example.
Looking at the post in general it seems that it's very much aimed at being as misleading as possible.
Then the blog post ends by saying CloudFlare is not the fastest in 10% of cases which seems simply false.
It says "we are aiming to improve our performance in 10% of the networks where we are not #1 today". That's not the same as "we're not the fastest in 10% of networks".
The network team is working right now with this data to improve in 10% of the networks and we're going to keep going.
I see, my bad. I guess I was a bit hasty since the metrics kind of bothered me.
Still, regarding the other point. Any thoughts on that? We've been monitoring performance of various CDNs ourselves, mainly to improve our own performance through our own RUM metrics as well as platforms such Citrix and PerfOps, and while CloudFlare is definitely one of the fastest, it's definitely not the fastest as per your results. For example CacheFly and Limelight seem to be excelling in the US way ahead of others.
Limelight being a multi-hundred million dollar revenue CDN was somehow excluded entirely.
Thanks, interesting paper!
But their technique only allows to measure if a resource loads or not, how does CF measure time to load on the end user browser?
12 comments
[ 6.1 ms ] story [ 17.7 ms ] threadIn fact I used to work with a huge adult site, where there were performance metrics down to a /24 subnet level - so in cases like that, performance problems could be investigated quite easily.
Then the blog post ends by saying CloudFlare is not the fastest in 10% of cases which seems simply false.
The metric selection as well seems to be selected very much in a way where it's skewing the actual results into making them be much better than they are. A much better result would be to actually show the actual global median response times, which would likely be extremely close to each other between all major providers. Meanwhile this post attempts to make CloudFlare look multiple X better than Akamai for example.
Looking at the post in general it seems that it's very much aimed at being as misleading as possible.
Maybe it's just me.
It says "we are aiming to improve our performance in 10% of the networks where we are not #1 today". That's not the same as "we're not the fastest in 10% of networks".
The network team is working right now with this data to improve in 10% of the networks and we're going to keep going.
Still, regarding the other point. Any thoughts on that? We've been monitoring performance of various CDNs ourselves, mainly to improve our own performance through our own RUM metrics as well as platforms such Citrix and PerfOps, and while CloudFlare is definitely one of the fastest, it's definitely not the fastest as per your results. For example CacheFly and Limelight seem to be excelling in the US way ahead of others.
Limelight being a multi-hundred million dollar revenue CDN was somehow excluded entirely.