So my understanding of cloud servers basically equates to shared resources, theoretically unlimited. Hence the facebook's and twitters of the world rely on cloud services.
Out of curiosity i'd like to know what you think causes espn3.com to fail so many times during the world cup.. the second a favored team scores, the site streaming goes down for myself and alot of friends. Sports discussion aside.. we have a media site refresh that gets good traffic will get even more once advertised. Obviously we're no world cup but I just thought that the idea behind the technology was that resources would be virtually limitless if you had the $$ behind it . And i'm guessing ESPN has the money..

Only they would know, but the cloud is only as good as the infrastructure it's built on, and the application that's running it. Not all things Cloud are equal.

Could also simply be their cloud is not large enough to support the streaming because they underestimated their required resources.

But again as cloudweb said: not all things cloud are equal.

clouds can handle the traffic that are designed to. Facebook has it's own cloud which in fact is a cluster of servers designed to handle facebook's traffic

stoidis, note that Twitter often has issues as well - hello fail whale ☺

This likely isn't a failure of "the cloud" in the sense you're thinking it is. This is likely a failure of their streaming / CDN architecture (which was recently switched over to MLB Advanced Media when they rebranded it from ESPN360.com to ESPN3.com), which is not all that uncommon for some visitors with these major streaming live events.

A world cup event is going to crash about any cloud there is. We're talking tens or hundreds of millions of concurrent unique visitors / streams added to the app overhead. I doubt anyone here has experience with such scale.

Regards
Joe / UNIXY

Thanks for the feedback . I know it's well.. the absolutely largest event in the world.. but it's also my personal test of the capacity / capabilities of the cloud concept. I can ONLY imagine how much traffic they are getting of course .. and the cost..

It depends very much of the infrastructure solution the ESPN cloud is using. I presume they went for an outsourced IT hosting service, so you'd blame on their host. When a serious company like ESPN comes to you, you'd come up with a few scenarios to make sure they will be up at any point.

But as you guys all know, thing happen and sometimes downtime can not be avoided.

A world cup event is going to crash about any cloud there is. We're talking tens or hundreds of millions of concurrent unique visitors / streams added to the app overhead. I doubt anyone here has experience with such scale.

Regards
Joe / UNIXY

BBC and Akamai managed something like 900,000 concurrent streaming users during one England match, without any problems.

The biggest problem during World Cup games was actually lack of end-user ISP network capacity.

The biggest problem during World Cup games was actually lack of end-user ISP network capacity.

That's my opinion too

Only they would know, but the cloud is only as good as the infrastructure it's built on, and the application that's running it. Not all things Cloud are equal.

No kidding. For all we know the back-end for espn3 was not built that well or the actual physical infrastructure wasn't that great.


Cheers,

Obviously ESPN would know the reasons in finer detail. In general though, the crashing of a cloud enabled high traffic site will have variables like: (a) the auto scaling software & algorithm
(b) the load fluctuation & its variance from expectation
(c) the application
(d) cloud infrastructure of course
(e) unwanted service requests

My 2 cents...

the cloud still runs on physical servers (called hosts). the amount of resources is finite, defined by how much total RAM, CPU, etc that the physical servers have. therefore, if enough users come on simultaneously, the cloud can eventually run out of resources.

Well a cloud environment is theoretically finite but can practically offer a lot of "computing elasticity".

As I said, there are more than one variables in the game.

Cheers!

even amazon runs out of nodes sometimes. I've heard that people often have difficulty if they want to get hundreds or thousands of nodes at once on amazon, and may have some of these requests fail. And I think by far amazon has one of the biggest and probably one of the most "elastic" clouds out there. If you've got enough capacity to handle even the most inconceivable load spikes, then you're wasting a ton of money on useless hardware, no question.

Also, though akamai could handle video here, someone else said espn3 was using a different CDN provider. It's entirely possible that the other CDN provider could not handle the load at some or all of their nodes. Also, some things only scale so far. If your database requires a single database server, that server can only get so big. It takes a lot of work to design a database to scale to an infinite number of machines.

Any network can be pushed to its brink, including a cloud. I wish my servers were being pushed to the brink as well :)

As others have mentioned there is a huge difference in the cloud providers out there. Some claim to be offering cloud but are simply running a hypervisor on a dedicated server - no SAN storage or shared resources at all. Be careful!