“Open” standards and your data centre

Blog

Categories

For more than a decade, the notion of ‘Open’ has assimilated nearly every corner of Information Technology. We can say that the beginning of the Open movement was marked with Linux. The open source community, believing that software should be free, consolidated creativity and labor across the planet to create a new operating system. That community changed the world, even if some Open contributions evolved into proprietary, commercial offerings. Once foreign and frightening to some, it is now a staple in the modern data centre.

Open sourced development spread rapidly into other infrastructure software and applications. The OpenStack project produced a complete cloud infrastructure stack including compute, storage, network and a control dashboard that can run on any commodity hardware.

Searching for new dimensions of efficiencies, hyperscale data centre operators began to recognise that there are great opportunities to optimise the computing resources they deploy for their applications. Instead of purchasing standard OEM servers, some began to challenge the way these off-the-shelf boxes are constructed, noting that they could be customised in more ways than simply selecting the number of cores, drives and amount of RAM.

This led to home grown, fully bespoke server configurations, selecting device-level components and assembling them optimised for the particular application they deliver. Facebook decided to leverage the open source community toward this end and created the OpenCompute project to advance this practice.

OpenCompute is not limited to the configuration of servers. OpenCompute has extended to include new specifications for servers, storage, and even equipment racks. With OpenCompute the vendors may be original design manufacturers (ODMs) rather than traditional original equipment manufacturers (OEMs). Where it was formerly the case that ODMs sold directly to OEMs like Dell and HP, now the ODMs claim server market share of their own because of the massive scale of data processing footprints used by hyperscale operators embracing OpenCompute.

The value of OpenCompute in this regard is economic. It facilitates commoditising and optimising the operational efficiency of the hardware used for computing in the data centre. These economies though, are realised at scale.

While we have things such as OpenStack and OpenCompute that are very data centre-centric, what shall be said about the network, which together with the data centre provide a holistic solution for the enterprise?

The Open Networking Foundation (ONF) organised to promote software defined networking (SDN), and supports the OpenFlow standard. OpenFlow is a network communications interface allowing direct access to the forwarding plane of network devices. This allows network control by locally managed software, rather than OEM-based platforms.

Projects such as OpenDaylight are advancing SDN and accelerating innovation and interoperability between otherwise proprietary platforms. The Open Platform for Network Function Virtualisation (OPNFV), similar to OpenDaylight is converging collaboration between industry players to create a carrier-grade reference platform for Network Function Virtualisation (NFV). Initiatives like these are facilitating network programmability and service agility into carrier products that can be procured by the enterprise.

With all this “openness” rampant across the IT footprint, what are the opportunities for the typical enterprise? Indeed the adoption of some open standards is prolific in the enterprise, in particular, around the software stack. Others though, are more difficult to reach. Let’s look at some reasons why this may be.

Capturing the economic advantage of some of these new open frameworks comes with certain prerequisites. Regarding OpenCompute for example, the ability to economically buy ODM equipment in bulk requires significant volume of purchasing.

Another factor deserving attention is the application mix. Hyperscale operators typically have a more homogeneous application inventory (or at least an application deployed massively at scale), often with multi-tenancy, that is delivered en mass to end users. Compare this to a typical enterprise, which is characterised by a heterogeneous mixture of COTS and bespoke applications with a wide range of hardware and stack configurations beneath them, as well as a variety of distinct user groups accessing them. Projecting open standards onto such environments requires consolidation in enterprise architecture.

The impact of organisational alignment is also not to be minimised. Many enterprise IT shops are still organised around horizontal tiers of infrastructure; network, distributed systems, storage. Many of these open models enable support vertically through the stack, and this can create management difficulties for the enterprise in both engineering and operations.

A further point is maturity of initiatives, technologies and products. While some open standards are de rigueur in enterprise environments, others are in the formative stages. The typical enterprise lacks the resources to push cutting edge transformations, and will wait for the time in which adoption is commonplace and the benefits are repeatable.

The innovation leveraged through the Open Source movement is impacting all areas of Information Technology, from the data centre floor through the carrier network, with the promise of interoperability, cost savings and operational efficiencies to be gained. These benefits are best realised in context however, requiring the end user to approach with planning and clear vision to properly gauge the benefits realisable in their environment.