Paul Hinks' Blog

Interesting news from GoGrid today – they are adding two new features which actually go some way towards addressing the cloud computing issues I have mentioned before. Firstly they are allowing user-defined images to be uploaded, from which multiple virtual servers with varying memory allotments; this means that I can clone my entire application stack and deploy multiple copies of it in minutes. They are also connecting their Colocation servers to their cloud servers, so it will be possible to use the standard hosting service and expand into the cloud as and when required, just as I said I wanted to here. This issue is also potentially addressed by Dynect’s solution, although I have yet to fully investigate that service. Anyhow, it’s good to see that somebody is thinking through the cloud offering and what it takes to make it a practical solution. As soon as I’ve fully investigated Dynect and kicked GoGrid’s tyres for deploying my new project I’ll post the details here for people to throw stones at.
[ad#co-1]

Further to my recent comments about cloud computing – I am currently helping a new site set itself up – we want to use a traditional hosting service but with the option of dynamically expanding into the cloud if bandwidth or other capacity limits are approached. Somebody had a bright idea – set up a free starter account a goggrid, configure a loadbalancer as the public address of the site and point the loadbalancer back at our server. Then when we need to we can start up a gogrid server and have the loadbalancer distribute load between the two servers. Brilliant – just the kind of scenario I wrote about, and a logical use of cloud capacity. Unfortunately gogrid don’t let you configure a loadbalancer to point to anything but their own servers. Rats. Something is going to have to give here if this concept is ever going to take flight.

I’m looking at doing some work for a cloud computing provider in San Francisco right now. It’s making me think a little bit more about the space than I have done before. I know everybody thinks the cloud is the “Next Big Thing” (I feel like there should be a copyright symbol after that phrase), and I do feel it makes a lot of sense to make computing power available in an on-demand, utility provider sort of model. The problem is is it really cost-effective versus the traditional hosted server model? And is it compatible with an in-house system that many larger companies run? Because is if it can’t work well with either of those two models then it really has no customer base.
I run the IT operations for a small company, and I know I want the most cost-effective model to host my servers. Right now that advantage is clearly with a traditional hosting company. I can see no way my application will suddenly require the extra processing power a cloud provider will theoretically be able to offer it, and in my case I see no easy way for the application to be able to take advantage of it if it was available. And the extra cost involved in this form of hosting negates any potential advantage it might offer anyway.
Since in my case, my core requirement is a traditional hosting service, I would expect that core purchase to be competitive with its traditional suppliers. If I do then require any surplus CPU power or storage space then I’ll pay cloud rates for that. If the new requirements become standard for me I’d expect to be able to add them to my core plan at a lower rate, and then only pay for resources beyond my new expanded programme.
I have some jobs that need to run for a couple of hours a month, and I’d love to be able to have those live out on the cloud and just pay for the time they spend processing – it would cost pennies a month from any of the providers. The trouble is they require storage of their state between runs, and that can’t be archived away with the job itself, so I’d again end up paying for resources that I’m not using.