Latency is a big issue for the app we're building [plug at the bottom]. Ideally we'd serve content in under 200ms to users around the world. I really can't see how to achieve this without either crazy Javascript hacks or Akamai. Looking up benchmarks suggests Akamai CDN has latency of around 10ms vs 100-200ms for everyone else (!!!) The killer feature, however, is Akamai's edge computing -- there are only two companies I know of that offer this. Unfortunately Akamai are stuck in the crappy business model where everything -- even basic information -- requires you to contact them. So we have to go through the bullshit salesperson theatre while they work out how much they can shake us down for. Not really what a small startup wants. So crazy JS hacks it will be, I expect.
Op plug:
App is http://mynaweb.com/
I just broke the sign up form so email us if you want in on the beta.
That's a fair assessment. Akamai isn't really ideal for startups (from a cost standpoint), but once a startup has gained traction in the market and is looking to scale (or optimize performance), you can't beat Akamai. Code optimization will only get you so far. Akamai's service will do things like find the shortest path (which cuts down on latency) and all sorts of other things upstream/downstream. Those things sound small, but have a huge impact to your site's performance.
I think my startup might have a good solution for you. That is if the content you are serving is of non-trivial size. We do not have servers around the world but instead work by improving the TCP protocol itself. Thus you do not have to worry about your server side software being loaded on different computers all over the world and how it would perform in various virtual machine environments, etc. Our software would be loaded on your servers only, and you can place your servers anywhere that is convenient for you.
We still have not officially launched but this is a barebones website:
extremetcp.com
Let me know what you think. You can reach me at
hristo at mainlinenet dot com.
Thanks for all the excellent comments! I should have noted that we're not serving static content -- we need to calculate a result for every request. The Akamai options I see are either to use their edge computing system, or to use ESI as a poor-mans routing algorithm (i.e. use ESI to include the IP address of our closest server in the Javascript we serve).
Among the more unique and impressive products are a set of route/path optimizers. They make constant speed, latency, congestion, and reliability measurements from thousands of points across the internet and use that data to send content through the most efficient paths possible.
Also, these services aren't just blanket across your domain or account - you can configure exactly what content gets what treatment based on literally hundreds of different variables. Furthermore, there are ways of having your origin servers send info to Akamai's edge caches that can then be used to make even smarter decisions, per-request.
I agree with other posters - it's expensive overkill for small sites or sites with modest needs, but for those who need maximum control, reliability and performance, I don't think there's any other company that provides it.
This paper is old. The media went running with numbers that are way outdated - The platform now has over 95,000 servers in 1,900 networks across 71 countries.
I don't know about ops but all of engineering is about 600 or so, IIRC. Everything is automated to an incredible degree though, so I wouldn't be surprised if ops is quite lean.
