Dale Vile, originally published on The Register
It sometimes seems as if the whole world has gone cloud crazy – well at least most of the vendors, pundits and many in the media. If we listen to the evangelists, the days of the enterprise data centre are numbered and players like Google, Amazon and Microsoft will inherit the earth. Even David Cameron, the illustrious leader of the opposition to our UK government, has been talking about handing over the country’s health records for storage and management to one of these big American multinationals.
In the midst of all this noise and hype, many have lost sight of the fact that getting a third party to run some of your infrastructure for you is a model that has been around for at least three decades. Indeed, those who have been taking advantage of hosted services, or on the other side of the fence, delivering them, must be wondering what all the fuss is about. Just what, exactly, is this cloud thing bringing to the party that’s apparently going to change the way everything works?
Well to deal with this question, we need to be very clear what we mean by cloud services. There are all kinds of definitions kicking about, but most of it boils down to three types of offering. If we start at the top, we have software as a service (SaaS), which is essentially based on the concept of renting application functionality from a service provider rather than buying, installing and running software yourself. Offerings within this range from services such as Salesforce.com at one end, delivering the equivalent of a complete application suite, to players like MessageLabs is at the other, who services are designed to complement your operational infrastructure.
If we drop down a level in the systems stack, we then have platform as a service (PaaS), which is all about providing, well, a platform in the cloud, upon which applications can be developed and executed. Players like Google, again Salesforce.com (this time with Force.com), and Microsoft (with Azure) exist in this space. Facilities provided include things like database management, security, workflow management, application serving, and so on. Once you have developed your application, it then takes advantage of the scalability and on-demand economics associated with the cloud computing model at execution time, which is great for highly dynamic externally facing applications, for example. The snag is, at least for today, that these environments are very proprietary, offering little in the way of portability should you want to move your application elsewhere.
We then get to the bottom of the stack, and it is here that we find the concept of infrastructure as a service (IaaS). The proposition here is the offering of compute power and storage space on demand. The difference between this and the other two categories of cloud is that the software that executes is essentially yours. In practical terms, the model is based on the same principles of virtualisation that we are all familiar with in the context of server partitioning or flexible storage. Rather than running a virtual image on a partition existing on a physical server in your data centre, you spin it up on a virtual machine that you have created in the cloud. Virtual disks can be created in a similar manner to deal with the storage side of things.
“Hold on”, you might say, “but haven’t we been doing this with traditional hosting companies and ISPs for a long time?” Well in terms of the basic principle of renting virtual servers from a provider to run some of your stuff on, then the answer is obviously “Yes”. The difference is that hitherto, we have generally been required to commit to that rental for a reasonable period of time and pay for the resources allocated whether we use them or not. In the cloud model, we can grab a virtual server for as little as a few hours then drop it again, and only pay for the time and/or resources that we have used. And if the load on the application fluctuates, the horsepower available to it can be adjusted accordingly. Rapid and automatic provisioning and deprovisioning, coupled with clever billing systems are the key to this on the service provider side.
With such an attractive model emerging, it begs the question of whether existing hosting companies and ISPs will be threatened by the new breed of cloud services offered by the likes of Amazon. To address this question we have to remember, however, that the kind of flexibility we have been talking about, or ‘elasticity’, as the cloud folk like to call it, is likely to come at a price. Indeed, as some of the traditional hosters have started to extend their service portfolios to include cloud type offerings, this difference becomes very apparent in their price lists (if you do the calculations).
Conveniently, however, while cloud computing advocates are constantly comparing the cost of IaaS with the cost of running on-premise equipment, they typically ignore the comparison between the elastic cloud model and the traditional static hosting approach. And when we consider that the majority of computing workloads running in the average data centre or even on existing hosted infrastructure are not that dynamic, this is quite an important difference to consider. After all, why would you pay a premium for elasticity and you don’t really need it.
Looking forward, it is clear that the on-demand IaaS model will find its place for executing applications that are genuinely very dynamic in nature, or for grabbing resources to deal with transient needs such as development, testing, one-off number crunching jobs, and so on. And many of the traditional hosters have already realised this and are moving in that direction. There will, however, continue to be a demand for the kinds of hosting we have always been familiar with. Whether it’s commodity server space, or more bespoke hosting arrangements, cloud is not going to kill the traditional approaches or providers.
In fact, as customers are likely to need a blend of static and dynamic hosting arrangements in most situations, providers who can offer a range of options within a single service portfolio, and for strategic hosting arrangements back that up with good quality support and account management, will continue to do well. Amazon will not, at the end of the day, rule the world.