This article is brought to you by ComputerworldUK in association with Intel IT Center
Progressive enterprises are currently investigating ways to optimise their use of public and private cloud technologies. They’re searching for an architecture that offers a clear roadmap to the future, and infrastructure and management technologies that can effectively manage hybrid cloud.
The answer lies in Software Defined Infrastructure (SDI), which has the potential to revolutionise the datacentre. It brings to the enterprise technology and techniques used by the ‘hyperscale’ service providers, companies such as Amazon Web Services and Google, as part of a hybrid cloud operation.
Hybrid Cloud is very attractive to enterprises that may have concerns about where data resides or who have substantial datacentres of their own. Like other cloud models, it offers lower operational and capital costs and improved agility, but it does more than this.
Hybrid cloud brings greater security and compliance, more granular data controls, and control of workloads, storage and network resources in a way that limits risk and increases efficiency.
Alan Priestley, Director, Strategic Marketing EMEA at Intel, comments that datacentres are traditionally built around the model of fixed workloads where there is a server for every application or process. Hybrid cloud, on the other hand, lets users make their own decisions about what runs where, and it provides datacentre managers with the orchestration and management tools to accomplish it.
The technology behind hybrid cloud is starting to enable organisations to seamlessly move workloads across providers, from on-premise to the cloud, and across different types of clouds.
This is where the hybrid approach really proves its worth because it delivers highly efficient workloads that are coordinated and optimised. It also means instant scalability through ‘cloudbursting’ – the ability to shift workloads onto large-capacity public cloud infrastructure to cater for peaks in customer demand, accelerate product testing, or fend off a DDoS attack, for example.
However, to create an effective hybrid cloud, it is important to have a holistic approach to datacentre transformation, to avoid the problems that can arise from having a piecemeal virtualisation or hybrid cloud deployment.
For example, there is a need to deploy the right workload to the right virtualised machine through effective orchestration on and off domain.
Also, there needs to be integration between the infrastructure and the application environment. In a hybrid cloud, the ability to spin up virtual machines for Infrastructure as a Service (IaaS) or combinations of IaaS and Platform as a Service (PaaS) must be the same in both private and public cloud environments.
As for the applications themselves, they must be portable. Hybrid cloud-aware development, therefore, builds capabilities into applications so they work the same across cloud environments.
In addition, there needs to be monitoring and management across cloud environments; with visibility into system health across clouds being critical.
Datacentre equipment manufacturers such as Intel have pursued a number of technology initiatives to overcome these types of issues.
The first of these is aimed at delivering an integrated system architecture across the cloud environment. Datacentre technology suppliers are working together on open standards and interoperability, to ensure that the combined architecture can support hosting and movement of workloads in a hybrid deployment.
The same is the case for infrastructure and application portability, as well as maintaining security, data privacy and compliance across cloud environments: these are now starting to be underpinned by multi-vendor cooperation and open standards.
At the processor level, Intel has built cloud technology features into its Intel Xeon E5 processor platform. Among these are Intel Turbo Boost, which accelerates processor performance for peak loads, automatically allowing processor cores to run faster than the rated operating frequency (if they’re operating below power, current, and temperature specification limits).
Another technology is Intel AVX (Intel Advanced Vector Extensions), which increases parallelism and throughput in floating point SIMD calculations because it expands most integer commands from 128 to 256 bits. Again, this helps with boosting processing throughput to deal with large workloads in a timely manner.
Intel AES-NI (Intel Advanced Encryption Standard New Instructions) can be used to strengthen cloud security, by using the processor to accelerate encryption and decryption across server and storage systems.
Another initiative is Intel TXT (Intel Platform Protection Technology with Trusted Execution Technology) again a processor-level technology that can be used to create trusted compute pools in cloud environments, addressing the issues of security, privacy and compliance.
Another exciting cloud datacentre development is Rack-Scale Architecture, which addresses the need for greater connectivity, real-time data, network efficiency and on-demand service delivery.
Intel and Ericsson have been working on a highly-scalable cloud infrastructure for the telecom industry: the Ericsson Cloud System, based on Intel Rack-Scale Architecture.
This implementation uses the Ericsson HDS 8000 server, which is built on a modular architecture with optical interconnects and disaggregated compute, storage, and networking components. This design enables resources to be better utilised, and has components that can easily be swapped out as technologies change.
In addition, being software-defined in nature, the architecture enables compute, storage, and network resources to be pooled, so workloads can be scheduled efficiently, and operators can increase overall compute density, utilisation, and network bandwidth. This resource-sharing architecture also means big reductions in power and cabling.
Lastly, by using this combination of disaggregated hardware and advanced infrastructure management software, datacentre managers have the ability to assign resources based on workload requirements, which can lead to significant savings in capital and operational expenses.
All of this is great news for the enterprise, which will either benefit directly from Rack-Scale architecture in their own private/hybrid clouds, or work with service providers who deploy it.
And moving forward, the explosion of mobile data, analytics and general web-scale activity means large enterprise datacentres will become de-facto Rack-Scale operations.
Consequently, before long, all enterprises will be seeking Google or Amazon Web services-type economies of scale in their datacentres, and hybrid-cloud, SDI solutions like Rack-Scale will provide the answers. Welcome to the datacentre of the future.