Servers, or instances as they are frequently known, represent the main compute resource in IaaS. Simply put, they will usually be instances of a virtual server running a standard operating system such as Windows or a Linux variant. The underlying virtualization technology used to support the servers isnt significant, although some service providers make a virtue of highlighting the hypervisor they use.
Operating system choice for servers will cover both Windows and Linux platforms -- the specific versions available will vary by provider. One point worth considering when choosing an O/S is the ubiquity of that platform across service providers.
Windows Server 2008 and CentOS are universally available (with Windows attracting an extra charge for licensing). Other variants of Linux are less popular. Amazons AWS takes operating system selection a step further by allowing the customer to choose from a range of AMIs or Amazon Machine Images. These include customized and pre-configured setups; currently there are over 7000 community customized AMIs to choose from.
When a new instance is created, the boot disk on which the instance runs can be persistent or transient. Persistent disks are retained when an instance is destroyed; transient disks last only as long as the instance itself.
Separating the instance from its boot disk is beneficial in a number of ways, as it can be replicated and moved around independently. Not all service providers offer persistent boot images and it is worth checking what features are offered in order to backup or snapshot the image in case recovery is needed.
The second feature is load balancing. A load balancer provides a virtual IP network connection and distributes connection requests across a number of instances of an application. This feature can be used to spread load across a number of server instances or to add a degree of resiliency and availability to an application.
For example, if a web application has periods of high demand, an additional server instance can be created and added to the load-balancing list for the duration of the increased demand. The temporary instance can then be decommissioned when demand subsides. The major IaaS providers all offer load balancing as a feature.
In addition to the two basic features discussed, some providers (notably Amazon) have a number of other offerings available. These include storage, database and messaging.
Poor physical controls can result in data breaches or worse, including prolonged outages. Logical security should ensure that unauthorized access cant be achieved in what is a multi-tenant environment. Always review the security features of your Cloud provider to ensure they meet your standards or compliance rules.
Application Programming Interfaces (APIs) enable Cloud computing to be integrated into existing business processes, including change control, provisioning and billing. And for organizations that already run their IT operations as a service to internal business customers, this will be mandatory. When multiple providers are used, APIs enable a common interface to be established, irrespective of where the computing resources are located.
Of course every provider will do differentiated pricing, including the costs of some services as part of the package and charging for others. Its worth ensuring you know the full details of what your cost model is and more importantly how that translates into any Service Level Agreements if the service is unavailable or performs poorly. This is probably the most important aspect of service provision to understand; your business could be affected by an outage against which you have no claim.