Datacentres around the world are increasingly hitting processing and connectivity constraints determined by limits in power, cooling and space. At the same time, capital expenditure spending restraints are forcing organisations to rethink their datacentre strategy. One avenue to explore involves converging infrastructure in a move away from siloed servers, storage, data and processes to more efficient use of the available power and space.
Converged infrastructure (CI) is still part concept, part smart marketing and part well-defined IT system components. It is variously described as: hyper-converged infrastructure, converged systems, unified computing, fabric-based computing, web-scale computing and dynamic infrastructure. Essentially, infrastructure convergence is about storage and computing systems that can integrate with hypervisor and networking technologies, providing the ability to abstract services.
Converged infrastructure born in the cloud
The CI concept originated in large cloud companies such as Amazon, Facebook and Google. These companies placed an emphasis on low cost, ease of use and rapid self-service. They tried to use existing infrastructure architectures, but quickly realised they were a poor match for their demands.
They wanted a much cheaper infrastructure to run any application at any scale. That required a software-defined system to break dependence on hardware for additional features and upgrades. As cloud companies they wanted agility, security and reliability of on-premise systems they could roll out globally.
In line with cloud thinking, converged datacentre infrastructure can serve as an enabling platform for private and public cloud computing services, be they infrastructure (IaaS), platform (PaaS) or software as a service (SaaS). Several characteristics suit on-premise CI to cloud deployments, pooling IT resources in a hybrid environment by automating resource provisioning. It also supports agile IT scalability to meet the needs of dynamic computing workloads.
For example, when launching a branch location with a few hundred users who need locally accessible resources for virtual desktop infrastructure (VDI) applications and a few hosted desktops, IT planners can adopt a CI approach that deploys a multi-node converged system to meet business demands. It can reduce deployment time by using a less expensive platform; and simplify management by unifying datacentre controls.
CI operates by grouping IT components (servers, data storage devices, networking equipment and software) into a single, optimised on-site package for infrastructure management, automation and orchestration. CI involves pooling computers, storage and networking resources, to be shared by a wider range of corporate applications; and managing these shared resources using policy-driven processes.
CI remains an evolving technology, with the ultimate aims of:
- Controlling with software all resources on top of the hypervisor;
- Application program interface (APIs) and virtual machine (VM) translation technologies allowing CI to be deployed on any hardware and interface with any hypervisor;
- Convergence software translating the VMs from one hypervisor to another using standardised underlying hardware to deploy hyperconvergence on blades, rack-mount servers etc;
- Cross-platform integrated automation and orchestration controls developing powerful next-generation datacentre and cloud models.
APIs integrating on-site CI with public and hybrid cloud platforms, integrating with other hypervisors and extending the corporate datacentre beyond a private architecture;
The corporate CI plan
When selecting a converged system, the level of automation associated with the management of the whole IT environment is a critical factor. Any investment in CI will have an impact on the corporate infrastructure for many years to come – so the IT needs to be adopted with a five- to 10-year investment timeframe in mind.
There are cost and performance advantages to be gained from converging the corporate infrastructure:
First, CI can lower capital expenditure and operational expenditure:
- Equipment and software life cycles are extended if the datacentre can increase server use, and reduce the amount of cabling and network connections;
- Lower operational expenditure comes from reduced staff levels when the network management infrastructure and datacentre management is automated, and storage is consolidated.
CI can also increase performance:
- Virtualising the networking for internet protocol and Fibre Channel storage;
- Allowing single-console management.
The CI market
CI comes in two main flavours:
- A single system with software and hardware components that are engineered together and identified as a single billable item in a company’s inventory. Oracle’s Exadata is virtually all-Oracle technology with very little flexibility from a technology choice standpoint. The advantage is like an iPhone – it’s all integrated. The disadvantage is that they limit choice in a very dynamic market. To provide end-to-end CI, hardware, software and services suppliers need to establish a host of integration agreements. But while this ensures their CI offerings can interoperate for the moment, it gives no guarantees as to how long the partners will go on developing their components in lock-step.
- A system that uses a reference architecture. This typically provides a list of functions and some indication of their interfaces (defined in APIs) and interactions with each other and with functions located outside of the scope of the reference architecture, notably cloud services. The reference architecture provides a template – often based on the generalisation of a set of real-world experiences – to identify CI components that have been observed in a number of successful implementations. It shows how to cobble these components together into a system. Adopting a reference architecture in an organisation accelerates delivery through the re-use of an effective system and provides a basis for governance to ensure the consistency and applicability of technology use in an organisation.
Reference architectures – which are proven and tested in the lab and often in the field – are becoming a popular option with the channel and many users who may specify their own reference architectures internally. Certainly this is the case with giants such as Google and Facebook but, increasingly, large commercial customers are creating their own reference architectures, working with primary suppliers.
The marketplace for hyperconverged hardware and software took off in 2011 and has significantly outpaced the growth of legacy, purpose-built infrastructure designed to support a single application. IDC estimates that overall spending on converged infrastructure will reach $17bn in 2017.
All the major infrastructure providers are moving into the CI market, either to attack other suppliers’ legacy installed bases or protect their own turf. There are single CI systems from HP ConvergedSystem, IBM PureFlex, Oracle ExaData, Cisco UCS, EMC VBlock, NetApp FlexPod, F5 Big-IP platform and Nutanix Xtreme Computing Platform.
On the reference architecture side, these come from: HP, Dell, SimpliVity hyperconverged data architecture, Microsoft and Intel’s Cloud Builders Reference Architecture.
Convergence among suppliers
Perhaps one of the best-known converged platforms is from VCE - which started out as a joint partnership between Cisco and EMC. VCE’s VBlock uses EMC storage, VMware compute virtualisation software and Cisco networking gear.
The various infrastructure components are architected to work seamlessly together. EMC later bought out Cisco’s stake in the company and so VCE became an EMC subsidiary. EMC itself is now itself transitioning to becoming a Dell company.
Besides adding VCE’s VxRack Node and VxRack System 1000 Flex to its own portfolio, Dell announced a CI reference architecture that integrates EMC’s rack-scale design with Dell’s PowerEdge servers and allows scalability into the thousands of nodes.
And besides this, the company recently announced several CI offerings of its own. The latest additions include Virtual SAN Ready Nodes for the VMware-based hyper-convergence systems, a Dell XC Series of hyper-converged appliances and hybrid cloud support for VMware’s Virtual SAN.
Balancing benefits and risks
The momentum CI has achieved over the past five years promises to simplify IT management tasks considerably, while simultaneously lowering equipment spend, shortening implementation time and reducing staff involvement. So it is well worth considering, given that several studies of real-world CI implementations demonstrate savings exceeding 20% in the first year alone – as exemplified by Canadian Empire Life Insurance.
However, CI is still being developed – and perhaps a few cautionary notes are required:
- CI may lower capital and operational costs – but what about business enablement and best-of-breed systems? Will a CI system provide the competitive edge the business is looking for?
- Be sure to identify the problems that need to be addressed and then derive measurable success criteria. When changing the datacentre paradigm, IT functions need to have clear key performance indicators to document their CI investment decisions.
- Identify both operational and application performance issues that you want CI to alleviate. Not all applications may be suited to CI architecture. Typically, in vertical industries such as financial trading and banking – where processing and transmission speeds are paramount to business success – to focus on convergence could risk losing competitive advantage. Similarly, applications that are extremely sensitive may not live up to the relevant governance requirements if they are located in a converged environment. So make sure to test core business applications in a CI environment before deployment.
- Understand the IT organisation’s IT maturity. Changes in computing, storage and networking affects the IT team structure. Are the right cross-issue competencies available? How much training/hiring is required?
Implementation best practice
Bringing CI into the corporate computing environment should not be a rip-and-replace event, but introduced as challenges are identified and locations need to come online. In an overburdened datacentre, it may be opportune to upgrade the platforms delivering commodity business applications – and green field deployments of VDI in branch locations may generate useful CI experience before adopting CI for core enterprise applications in the datacentre.
Read more about converged systems
We offer some tips for IT leaders planning to consolidate their IT components into a single, centrally managed platform.
As hyper-converged infrastructure continues to gain momentum, vendors have raced to build their own platforms, sometimes at the cost of quality.
Bernt Ostergaard is service director at analyst Quocirca