Hacker News Reader

Comment by patrickgzill | Hacker News Reader

What is the barrier to entry for CloudFlare? Putting a few servers in different colos, buying quality bandwidth, set up BGP anycast, get the IP Geolocation database from MaxMind. We are talking a service that could be duplicated with open source software and about $100K to $250K investment.

View on HN

"We are talking a service that could be duplicated with open source software and about $100K to $250K investment."

On the other hand, that is the baseline pitch that has been used by starry-eyed start-ups for ages now.

One of the least informed opinions I've seen on HN.

Enlighten me... a 1U server can handle nearly 1Gbps of traffic. BGP anycast - a solved problem. How many different locations do you need? For USA - Chicago, NYC, WashDC, Boston, LA, Phoenix, Dallas. In Europe - Amsterdam, Frankfurt, Berlin, London, Paris, Moscow/StPete, probably a few I have missed. Australia, Singapore, Taiwan, HongKong. Half a rack (20U) in each location would more than cover most, though possibly the USA would require more.

http://fastly.com does that: Varnish on SSDs.

But Cloudflare isn't just a CDN, it has an application firewall/security thing built in too.

That's important because suddenly you can't get away with just Varnish, and rolling Apache+mod_security out to each node is probably insufficient as well.

Edit: fixed url, thanks.

http://fastly.com

CloudFlare also provides basic analytics, DNS, url rewrite rules, and a fairly decent UI.

The idea of rebuilding it for $250k is ludicrous.

And fastly is crazy expensive.

Try it for $250K ...let me know how that goes.

Reply on news.ycombinator.com