24
Jan
2010
The Changing Face of Infrastructure Part 3
The first post in this series I discussed the employee vs employer owned workstation (laptop) and the centralised vs distributed VDI infrastructure. In the second I spoke about Server Virtualisation in the Enterprise. This post I will take up where I left off and discuss the Private Cloud.
Cloud computing is not a new concept, it is simply a new way of looking at existing infrastructure. To start I will define what I mean when I discuss these concepts.
Firstly the “Cloud” is a highly available infrastructure which can either be used to host services, infrastructure or software either for a private Enterprise or as a public hosting platform.
The “Private Cloud” is a Cloud for use by an Enterprise solely for the purpose of running their Infrastructure and Software. Whilst some servers hosted in the Private Cloud may offer external services, such as a corporate Internet site, FTP, SharePoint or Business 2 Business portal, the Infrastructure is owned and run for the sole purpose of the Enterprise.
The “Public Cloud” is a hosted platform which provides services to customers. An example of a Public Cloud is Amazon’s EC2 (Infrastructure as a Service) , Microsoft BPOS (Software as a Service) or even Gmail (Application as a service). Each of these offerings provide a different layers of compute to the end user.
A Private Cloud is really just new terminology to refer to a dynamic datacentre. The biggest challenge with cloud computing is the management of the Infrastructure hosting the platform and the applications, services and servers which utilise it.
The idea of a cloud is that it is highly available and the only way to provide high availability is to distribute the Infrastructure across multiple sites to remove Single Points of Failure (SPOFs). Ideally a cloud should have no SPOF’s however this isn’t possible because at some point the Earth becomes that SPOF. Generally most SPOF’s can be removed by simply increasing the distance between datacentres, however be aware of International Borders, and Earthquake Fault Lines can sometimes add a level of complexity.
Inside the datacentre Server Virtualisation is the key, I covered the benefits of virtualisation of Part 2 and so won’t cover it again here. Whilst Server Virtualisation is the corner stone of a private cloud, for the cloud to be dynamic many other technologies need to introduced.
The first technology is monitoring. Monitoring is more then just a simple is a server up or down. All monitoring applications will also provide details about services, processes, monitor performance and even alert based of event log entries. However the real benefit of monitoring is the ability to recognise trends in Infrastructure health and performance and use the data to proactively resolve issues before they cause outages. The data can also be used as a trigger to further investment in Infrastructure as the existing server resources and storage are utilised. In the world of public cloud providers monitoring is also tied into the billing system where the user pays for the clock cycles they use.
Management of the Infrastructure needs to be flexible and dynamic and allow for users with differing levels of Administrative control. For the datacentre Administrator the ability to create, destroy, power on / off, move VM’s between hosts etc. needs to be a seamless activity that can be controlled interactively or scripted and scheduled as required. For end users there maybe a requirement for a provisioning portal to spin up and destroy VM’s as required for projects. This ability though adds more complexity again.
Once a server has been spun up in the cloud the virtual instance needs to be managed though lifecycle. It needs to be licensed, backed up, have the correct anti virus settings applied, monitored and patched, domain joined and basically managed like any other server in the environment. All of these activities need to occur with the little or no manual work by the Datacentre Administrator.
Having discussed at a high level the idea of a cloud and some of the technologies required for a private cloud it becomes clear that the biggest requirement when building a cloud isn’t technology but process. Without process and strategy many Organisations suffer from server sprawl as spinning up a new server no longer requires a CAPEX.
In the next post I will dive into a few of the products that I’ve used and discuss how their use can increase the ROI of a private could.