Can your cloud balance supply and demand?
The perfect match
Private cloud involves a multi-layered approach to architecting IT systems and delivering services to the business, making the most of virtualisation to provide a separation between the two.
There is no clever magic involved, nor does there have to be anything particularly special about the servers, storage and networking equipment used. Rather, it is a question of arranging resources in the right way and using management software to make the most of them.
To understand how things need to look, we can consider private cloud in terms of supply-side capabilities that consider physical assets and demand-side capabilities that focus on how services are delivered.
Do the automation
The trick, if there is one, is to ensure that supply-side capabilities are adequate, and to use software automation to enable these to be provisioned and managed in the most efficient way.
Underlying both supply and demand is the requirement for resilience to be built into the infrastructure.
Private cloud adoption involves consolidating multiple applications onto a reduced hardware footprint, which therefore increases the risk of failure.
If power goes down to a single server running a single application, for example, only one application is lost. But if a server is running multiple virtual machines, the damage could be far greater.
So it is important to ensure adequate provision is made for redundancy and failover across the architecture.
Operational management of supply-side capabilities requires a clear understanding of the server, storage and networking infrastructure and how it is packaged into logical units.
To an extent this equates to good, honest, traditional IT operations – configuration management of equipment, components and firmware versions, patching status and so on.
The fact that the environment is dynamically configurable acts in IT operations’ favour. It should be possible to move workloads to other areas of the infrastructure without incurring downtime, though this will not always be practical.
Which brings us to capacity planning. This can to some extent take place in advance, for example through a review of facilities and forward projections of processing, network bandwidth and data storage growth.
But the ability to monitor resource utilisation across the private cloud is an essential prerequisite to future planning and procurement.
Capacity planning at a physical infrastructure level also enables a picture to be built of virtual capacity, for example in terms of number of virtual machines that can be supported. Creation of sufficient supply goes hand in hand with demand management. The first step is building an understanding of how logical workloads map onto physical assets.
For example, how many virtual machines, in what configuration, are required to support a messaging server or a database application? Do these need to run on the same or on different parts of the physical architecture for redundancy? How should such facilities scale in case of increased demand?
Such information can be incorporated into policies governing how to create logical workloads, then feed into the provisioning process, that is to create a specific instance of a logical workload to respond to a specific need.
The business is less interested in how a service is provided than in making sure it will deliver on expectations, so this is also the level at which service level agreement criteria can be built in and coded into policy terms.
It is not hard to see how conflicts may arise. Even well-planned environments do not have unlimited resources, so trade-offs will have to be made – for example between completing that urgent analytics job or allocating resources for software testing.
Perhaps both needs can be met, but maybe not at the same time as the monthly reporting cycle… and so it goes on.
In response to this problem, the two weapons private cloud infrastructure managers have in their armoury are monitoring and automation.
- Monitoring because request fulfilment can only take place with a clear idea of what resources are available.
- Automation because it is impossible to allocate logical resources efficiently without scripted mechanisms to run up and configure virtual machines appropriately.
Above supply and demand, higher levels of management oversight are feasible – for example, linking IT service delivery with business service delivery and flexing IT services dynamically with the needs of the business.
However, this depends on having a resilient private cloud infrastructure that can deliver on an organisation’s needs with confidence.
For many, arriving at this point is already a major step forward. ®