Fuelled by increased desires for greater efficiencies, supportability, extensibility and cost effectiveness in IT infrastructures, Cloud computing has acquired an almost mythical status as a powerful consolidated architecture for delivery of enterprise IT services. Despite its close association with long-established technologies like virtualisation, grid computing and clustering, the concepts behind Cloud implementations are still fairly young. Notwithstanding this newness, businesses are subscribing to the model in increasing numbers.
According to EMA research, Cloud computing services are already a $40 to $95 billion dollar industry. Industry observers expect this to quadruple to $160 billion by 2015 and yet it is clear that Cloud computing is still in its early-adopter phase. EMA survey results reveal just 16 percent of organisations have as yet considered Cloud solutions for their IT services, indicating a wide potential for future growth as the value proposition becomes more widely accepted. Building a responsible Cloud infrastructure - either from existing computing resources or from the ground up – requires three important steps at the design stage.
Step 1. Effectively Size the Infrastructure
The first objective must be to support the maximum amount of computing services with the minimum number of computing resources. Capital expenditure on brand new environments can be reduced greatly by investing in a smaller number of more powerful resources. Meanwhile established environments can utilise existing systems more efficiently through consolidation efforts or by maximising system utilisation with grid computing. The fewer physical components there are in an infrastructure, the fewer systems there are that must be deployed, configured and maintained.
The second aim must be to reduce the number of potential security and failure points and improve overall reliability. Consolidated servers are also more easily pooled and can better accommodate the use of shared services, such as with commonly accessible storage and clustering. This allows new resources to be more rapidly provisioned to meet service demands. For example, as requirements for Cloud services grow, new disk drives can easily be added to an existing storage pool and new servers can be easily introduced to a cluster without affecting service availability.
Capacity planning is critical to effective infrastructure sizing. Systems that consolidate a large number of resources, such as blade servers and mainframes, are physically larger and more powerful than standard servers. Organisations must ensure data centres do not exceed environmental conditions – such as weight restrictions, power availability, and network bandwidth – and that sufficient environmental resource are always available to support rapid expansion. Automated tools can help track system and environmental resources availability so that quick and informed decisions can be made on the most effective ways to size and expand Cloud services to meet an expected growth in service requirements.
Step 2. Ensure high Availability
Since Cloud services are expected to be continuously accessible, high availability is another critical infrastructure requirement that needs to be addressed. The virtues of clustered servers have already been mentioned for their ability to be expanded rapidly, but their chief value comes from their ability to load balance system resources across cluster members and to provide uninterrupted fail-over services in the event of a single server being subject to catastrophic failure or maintenance downtime.
Clustered environments are typically contained within a single physical location so that they can share storage systems and do not have any performance latency due to WAN traffic. Large Cloud implementations typically have multiple clustered environments at multiple facilities. This allows failover of a Cloud service in the event of a site disaster. Individual Cloud deployments can be expanded to operate across multiple clustered environments, both local and remote, to create a “hub and spoke” architecture that ensures highly available and highly reliable compute services. Even by eliminating the possibility of a single point of failure, however, catastrophic system or site failure can impact performance of Cloud services, so automated tools should be employed to monitor the health of these systems as well as the availability of support services, such as power and network connectivity.
Step 3. Minimise Operating Expenses
A core expected value from Cloud adoption is the reduction of IT operating expenses. In fact, EMA research has confirmed that roughly 76 percent of Cloud implementations have resulted in significant and measurable cost reductions. Consolidation efforts can certainly help to reduce capital expenditures, but with a little extra planning and consideration operational cost reductions can also be achieved. Administrative staff, for instance, can be reduced and yet still sufficiently sized to meet support requirements by simplifying the environment to be managed.
Unlike traditional data centre models, Cloud infrastructures provide the opportunity to standardise on only one or a few system architectures. The fewer server, configuration, and application types there are to support, the less specialised knowledge is needed to support staff pool. Similarly, standardising management practices and automating procedures reduce the amount of administrative tasks and effort that must be performed. Monitoring and asset discovery tools can also reduce the administrative burden by making it easier and faster to identify root causes of problems and to proactively prevent potential problems.
Cost efficiencies can also be achieved by reducing data centre energy consumption. EMA research indicates that, on average, more than 20 percent of data centre electricity costs can be eliminated with basic power management practices, and more than 80 percent reduction in consumption has been achieved by the most energy-efficient data centre reconfigurations. The most effective server level power management practices involve monitoring the environment to look for opportunities to reduce consumption during low use periods. Significant cost saving can also be achieved by reducing data centre cooling requirements. By identifying “hot spots,” data centres can be reconfigured to ensure proper airflow and heat dissipation so that chillers do not need to be run more often than necessary.
Conclusion
The foundation for end-to-end Cloud infrastructure intelligence is being established today. Visual modelling tools have the potential to revolutionise the concepts of IT management. By transitioning from a passive monitoring solution to an active management interface, the solution can be transformed into a centralised console for unified service management. In time such tools will be integrated with more extensible automated systems management processes for things like provisioning, configuration, security, patching, and remediation services. The combination will result in a very intuitive virtual data centre representation where individual services can be modelled at the process level as well as at the physical hardware level.
