As enterprises look to increase the performance and flexibility of their core IT and data assets while streamlining expenditure into an OPEX model, how do they evolve from physical to hybrid and virtual datacentre environments and embrace cloud-orientated architectures that are progressively more web-scale and software-defined?
Legacy IT architecture typically follows an aggregation model. Layers of infrastructure consolidate traffic from the edge of the network into a core datacentre which hosts the processing and then passes the results back down the chain to the service user. The traffic is very 'North-South' oriented.
The massive infrastructures of cloud service providers like AWS, Microsoft Azure, Google and Facebook generally follow a much flatter topology that is far more scalable and manageable. Applications are spread across clusters of servers which are mainly in a single layer and hence traffic is far more 'East-West' oriented.
These cloud based datacentre architectures are highly elastic and scalable, readily enabling more compute and storage resources to be added or removed from a workload without significant networking changes. They commonly feature extensive use of virtualisation to maximise compute/storage utilisation and to minimise power and cooling requirements, also leveraging provisioning and configuration automation as well as self-administration. All this means faster, more agile delivery of new services to the enterprise employee end-user.
Cloud based datacentres also benefit from the extensive use of commercial-off-the-shelf hardware and open source OS, and this is very different from the expensive single vendor integrated appliances often found in enterprise datacentres.
By capitalising on these same technology principles, enterprises can rewrite the economics of their IT infrastructure and move to a hyper-converged, software-defined, webscale datacentre platform with extraordinary flexibility and responsiveness to help them foster innovation in their IT department and place their business at a distinct competitive advantage.
The main point with infrastructure transition, however, is that it must be done within the context of an application. Enterprise IT departments can't simply assume that a horizontally scaled infrastructure will entirely match their requirements. Applications based on a relational database management system might scale more easily simply by increasing CPU speed or the amount of cache (vertical scaling – scale up) whereas a web oriented application might scale better through the addition of more servers (horizontal scaling – scale out).
Most enterprise IT departments have already experimented with or deployed server virtualisation. The opportunity is to extend the concept of virtualisation throughout the IT infrastructure, fully embracing storage and networking to deliver a 'virtual datacentre' which is so flexible that it can be arranged into blocks of compute, storage and networking. Applications no longer need to have their own dedicated and attuned infrastructure. Instead the infrastructure is common throughout, but each application is allocated the appropriate amount of resources to deliver the desired performance. The resulting homogenous enterprise infrastructure enables larger volume savings, and dramatically improved resource utilisation and real estate reduction - in turn reducing power and cooling demands.
And because it's the same infrastructure, uptime can be achieved in a more cost-effective manner by virtue of high service availability delivered by moving workloads onto virtual machines on another physical server. Whereas applications were often built on dedicated and expensive redundant hardware, now backup servers can be shared amongst a variety of applications in 'N+1' fashion.
But virtualisation is not completely without some downside. In some enterprises the ease of server virtualisation is creating 'sprawl' because it's just too easy to spin up another server (e.g. to remove an application bottleneck) and then forget about it once the immediate congestion has dissipated. The same can happen to virtual storage.
The real pitfall of server and storage virtualisation lies in network routing. Moving workloads around a virtual machine architecture spread across multiple racks of servers can be 'mind boggling' in terms of routing. Important consideration needs to be given to network performance and this is driving the market for 'Software Defined Networking' (SDN) and 'Network Function Virtualisation' (NFV).
Bringing flexibility, scalability and agility to enterprise ICT infrastructure concerns the transition of applications into a virtual and converged computing-storage-networking environment, but also involves the automation of provisioning, monitoring, management and reporting. With all applications running in a virtual environment on a homogenous infrastructure, an immediate benefit will be the reduction in admin terminals needed for support purposes – bringing the goal of a 'single pane of glass' closer to reality.
As enterprises seek to gain the flexibility and agility of a public cloud via an entirely virtual IT infrastructure made up of both public and private cloud environments, the role of the IT department will transform as much as the datacentre. Rather than consume so much time and effort tending to the needs of infrastructure, they will become far more focused on the evaluation, selection and delivery of services for the benefit of their users.
Jason Dance is managing director of Big Technology UK