DISQUS

skwp · 1 year ago

Good point James. I don't think clouds remove the need for capacity planning. You would still want good predictive capabilities as well as anticipation. Most companies know when they're going to be featured on TV or on a major blog and can prepare in advance. Very rarely does a 100% traffic spike occur without prior knowledge. Clouds are letting us provision on demand, but we still need to tell them when the spikes are coming :-)

jamesurquhart · 1 year ago

I'm not sure I agree that capacity planning is easy for a blogger or news provider to do. Ask John Williams about the sudden interest in his post about "captchas". Mosso seemed to handle the surge, but I wonder if he lost a few people getting to peak capacity.

Even a SaaS offering could come up with a feature that unexpectedly draws the attention of a large audience web site. In any case, however, the more that is known up front, the more responsibility it is for the target site to prepare and/or warn its cloud provider.

Anonymous Coward · 1 year ago

"Schlossnagle" - what an awesome name!

ElasticHosts · 1 year ago

James identifies a real issue, but I do not see the picture as bleak as he does. A user of a grid service such as ElasticHosts or the others has two fallbacks:

1) For the initial response to the spike, the user can indeed set up with "warm" capacity - and this works very well in a cloud computing environment. e.g. they can size their virtual servers for large CPU/RAM with a low base utilization - while they have low usage the hypervisor will share capacity with other VMs, but it is instantly available when the spike hits.

2) Although the spike hits instantly, the higher level of traffic lasts for hours afterwards. With a good monitoring and a good cloud vendor, the user will be aware and able to provision additional capacity within the first hour. This doesn't tackle the initial part of the spike (handled above), but does mean that many of the NYT/Digg readers will actually get a great experience when they arrive later in the day.

Roland Dobbins · 1 year ago

The problem is that you've conflated what's essentially a server farm composed of VMs with true cloud computing. A true cloud isn't going to 'provision more servers', it's simply going to open more TCP sockets, HTTP listeners, whatever.

Now, the resources which power the cloud may well consist of VMs (like EC2); but the cloud isn't the VMs, it's the common code running amongst them which essentially transforms them into a single logical entity.

jamesurquhart · 1 year ago

So, Roland, you are essentially saying that Amazon is not a true cloud? Last time I checked, Amazon could only scale a server instance as much as a single physical box, after which you would be required to spin up an additional instance. Oh, and Amazon doesn't supply any built in monitoring or automation, so you have to patch that together yourself.

Same seems true of FlexScale, Mosso and others called clouds. Do these not fit the definition of a true cloud?

Roland Dobbins · 1 year ago

I don't think they've arrived at the true cloud yet - I think they're in a transitional state, and folks such as Amazon, FlexScale, Mosso, et. al. are providing the underlying infrastructure for what will evolve into clouds. My view of a cloud is that I can just write my code and fling it into the cloud, along with some metadata specifying resource allocation parameters, and I'm off to the races.

jamesurquhart · 1 year ago

While I agree that is the ideal vision in many cases, I reserve judgment about whether that is the ultimate definition of cloud computing. Its not that I don't share the excitement about that level of dynamic computing, its just that I am not sure that it applies to all cloud-based applications, nor do I believe you can ignore the problems that I raise during this "transition phase".

Good discussion. You have made me think a bit, which I always appreciate.

Roland Dobbins · 1 year ago

Oh, I agree with you that folks need to consider the issues you raised; as other have pointed out, there are ways of dealing with this, albeit imperfectly. Your points are well-taken, which is one of the reasons I subscribe to your weblog - you're bringing up valid issues.

In my mind, 3Tera are closest to enabllng what I consider to be a 'true cloud'. Right now, Amazon, Mosso, Terremark, FlexScale, Slicehost, et. al. allow customers to essentially build their own 'virtual private clouds' atop VMs running on a common infrastructure; 3Tera seem to be offering the whole enchilada, or at least more than the others, at present. And with full instrumentation, too.

Liftchair · 9 months ago

I'm sure that advances will appear over the coming years to bring us closer, but at the moment there are too many issues and costs with network traffic and data movements to allow it to happen for all but select processor intensive applications, such as image rendering and finite modelling.