Why scale out matters
Beware the credit card
It’s just a little piece of plastic. Yet any part of your organisation can use it to buy infrastructure on demand, in a public cloud, right now. And that’s dangerous. Say marketing want to run a campaign, and they have a rough idea of the IT resources they’ll need. In this time of IaaS, PaaS and SaaS, the price goes with the usage, so the cost of resources will be variable. It’s true that variable cost and speed of execution is what every project manager dreams of — but the danger is that your own IT department is nowhere near as fast and flexible. Traditional projects require careful planning, return on investment calculations and months of processing, and in some cases they result in over-buying and poor service. Even your accounting requires a lot of IT resources, with spikes in demand at the end of the months, quarters and financial years, and idle lulls in between. Turning IT expenses into a variable cost is what project managers want. The pressure then comes to transform your IT department into an on-demand elastic infrastructure: a real cloud.
Everyone in IT is putting “cloud” stickers on their services as soon as networking is involved, and the temptation is to ignore most of it as hype. But there’s a profound change coming that you won’t want to miss. Elastic, on-demand infrastructure is relevant and real enough to survive the hype. You can publicly buy scalable infrastructure today on Rackspace, hpcloud.com, Amazon AWS, Microsoft Azure and many others, but you can also have it on your own datacentre thanks to OpenStack. (Sure, there are others too, but OpenStack is backed by Canonical and I’m biased!) The cloud industry calls an Infrastructure as a Service (IaaS) in your datacentre a private cloud, and calls an outsourced one a public cloud. Those terms have nothing to do with the public or private access of the data and services that will run on them. Putting your old virtual machine with a monolithic service onto a “public cloud” is simply outsourcing and probably insecure. There’s nothing “on demand” or elastic there. If the only benefit you can foresee for moving to the cloud is “not running the service yourself”, it’s just disguised outsourcing. It might still have some benefit, as outsourcing is something you know and understand already. But this creates confusion and distracts from the real, profound change occurring today towards on-demand elastic infrastructure.
Small is beautiful
How can you tell if a vendor is “cloud washing” an old outsourcing story or, even worse, an application not designed for the cloud? First, follow the money. Does the price increase and decrease with usage? Is the data separate from the application? How different is the cloud version from the previous non-cloud version? If the service can’t scale up and down with demand, it’s not elastic. If the service can’t scale down to “tiny” instances, it probably wasn’t designed for the cloud. Another important aspect is resiliency. Can the service survive a virtual machine crash? Has it been tested for this? As an example, Netflix pushes its testing as far as randomly crashing resources to ensure the platform stays up. The program that introduces these random failures is called “the chaos monkey”.
Cloud services cheat sheet
To check for cloud readiness, answer these five questions:
- Does it start tiny?
- Does it offer variable “as you go” usage pricing?
- Has it been designed for resilience and elasticity?
- Are applications and data separate?
- Is the messaging clear? Public or private is nothing new; think owned or outsourced.
Adjusting the organization
Project managers should be able to understand elastic services, as should architects and consultants. Don’t be tempted to teach an old dog new tricks: if it’s not elastic it can “run” on cloud infrastructure, but it won’t benefit from it — and you can reach a point where it will be more costly and less effective to do so. IT departments must provide teams with an elastic, on-demand infrastructure where the organization still owns the data. They need strong project management and resource monitoring, and be able to anticipate demand over the weeks it takes to install new hardware, scrap obsolete ones and price their services. Few traditional IT departments are ready for this, which is the most challenging part of moving to your own “private” cloud, but the overall benefits are enormous in term of efficiency and time to market. More outsourcing should be beneficial; it’s unmanaged outsourcing that’s the issue. Cloud or no cloud, you decide what to outsource. Once the cloud hype dies down, on-demand scale out is here to stay, and it has the potential to change your organisation for good.
Embrace and flourish
By now you should have a better idea of what’s at stake, what kind of services to run on a public and private cloud, and how the cloud can help your business. The change is so radical that most of today’s applications will have to be redesigned to benefit from IaaS. This cloud will give businesses a completely variable, resilient and elastic way of delivering IT services. This journey has seen a lot of new companies, services and applications emerge, which means more choices and resistance from the incumbents. The ability to take more risks can unleash your workforce’s creativity and increase innovation.
Ubuntu offers all the training, software infrastructure, tools, services and support you need for your public and private clouds.