Simon Bearne, Commercial Director, NGD
As featured in Networks Europe - Cloud Computing – March 2018 (page 22)
Cloud services can be expected to deliver on the IT requirements of thousands, even tens of thousands of organisations.
Raising the stakes even further, users increasingly demand hybrid solutions that give them the best of all worlds by being able to seamlessly interconnect private and public clouds together. Many user organisations also have a practical need to encompass legacy IT workloads which can operate alongside the hybrid cloud environment. Further, there’s a growing trend towards multi-tenanted private clouds being offered by some service providers.
These developments are placing even more onus on the resilience and efficiency of the data centres behind such complex and demanding cloud structures.
Some of the major public cloud providers are taking steps to make the development and deployment of hybrid solutions more straightforward. The newly launched Microsoft Azure Stack, for example, is intended to allow organisations to run Microsoft Azure IaaS and PaaS services directly within their own data centres, whether in-house or in their chosen colocation facility.
On paper, this allows them to enjoy the full range of public Microsoft Azure services on their own hardware, while also moving private workloads seamlessly between their chosen data centre and the Azure public cloud. The major advantages here are continued ownership of core and mission critical applications in a private cloud while also receiving the added benefits of continuous software updating and automated backups delivered with Azure public cloud service.
Such initiatives are clearly essential for getting hybrid clouds well and truly off the ground. There are many organisations out there, especially more heavily regulated ones, demanding the retention of private cloud infrastructures and certain legacy systems which cannot be moved into a cloud based infrastructure. An organisation might be happy enough using an Internet-based public cloud development platform for testing new applications, but not once it goes into production due to security, latency and privacy issues.
In practice, whether in house, collocated or both, the need for modern data centre infrastructures is inescapable, capable of supporting the exponential growth of cloud environments today and in five, 10 or even 15 years’ time. With this, power and connectivity are absolute priorities.
Historically the electricity required by IT equipment has trended upwards but the huge demand for cloud services is accelerating this dramatically as well as the amount of rack space required. There is a common misconception that running low density racks instead of higher density ones will be less costly when it comes to power but the reverse is actually the case.
Running fewer high density racks than lower density ones will yield a lower total cost of ownership because they have far superior compute capabilities while using significantly less data centre resource; switchgear, UPS, power, cooling towers and pumps, chillers, lighting and so on.
Therefore, it’s increasingly important that a data centre provider designs their facility to accommodate high density racks and can achieve the right balance between rack space and power. Many racks installed in data centres now consume more than 10kW, and some even 60kW. Few can supply this level of power per rack today and this problem is only going to get worse.
Data centres are becoming trapped in a ‘perfect storm’ of rising demand for rack space and more power required per rack. This is why larger and hyperscale datacentres with more abundant power and space are inherently more suited to meeting and future-proofing larger enterprise cloud hosting requirements and those of the major public cloud service providers.
Facilities which are less dependent on the electricity distribution network, where the bulk of electricity supply failures occur, or better still, connect directly to the National Super Grid, are also likely to benefit from far greater reliability and fewer capacity limitations.
At the same time, as demand for hybrid cloud environments continues to grow, data centres must meet user expectations for application responsiveness and predictability. With the considerable amounts of data moving back and forth between the public and private cloud environments, and possibly legacy systems, a hybrid approach brings both latency considerations and the cost of connectivity sharply into focus.
Taking Microsoft Azure Stack as a working example, it can be run as a standalone or in a hybrid infrastructure alongside Microsoft Azure Public Cloud as a peer system. In the latter case latencies between the Azure Stack system and the Azure Public Cloud will determine how fast and seamless a hybrid cloud system is once deployed.
However, few private data centres will be able to afford to run the dedicated network links necessary for assuring consistent performance on an ongoing basis for workloads that may have variable resource needs. While for ‘standard’ interlinks between existing Microsoft environments and Azure Public Cloud, Microsoft offers ExpressRoute as a low-latency dedicated connection, it is only available as a trunk connection to certain colocation, public cloud and connectivity operators. These can connect directly with ExpressRoute at core data centre speeds and so largely eliminate latency issues, improve security and ensure bandwidth is optimised and predictable.
For those organisations not using private or colocation data centres directly connected to Microsoft ExpressRoute, the only alternative is to setup a fast and predictable connection from their facility to an ExpressRoute partner end point. This means that there are two ‘hops’ to get to Azure rather than one so doubling the costs as well as the possibility of additional network problems. This is the case even where connectivity providers are offering ExpressRoute to a private or colocation facility as they are layering their own connectivity from the edge of their network and the ExpressRoute core to the edge of the user network.
In addition, if an organisation is planning on using a colocation facility for hosting some or all the hybrid cloud environment but keeping legacy workloads operating in its own data centre, the colo provider must offer a range of diverse connectivity options. Multiple connections running in and out of the facility will assure maximum performance and resilience.
Last but not least cloud hosting data centres also demand highly skilled engineering personnel on site. Hybrid clouds are complex and cannot be built, tested and managed successfully without suitable facilities and training. Further, studies usually show that the majority of outages are caused by human error. Engineers must be well trained, and critically, know when to intervene and when to allow the automated systems to do their job.
In summary, the major cloud and data centre providers are working hard to meet growing demand for all flavours of cloud solutions. There are many efficiency benefits to be realised when moving to a cloud infrastructure, however, care must be taken to ensure that the best location for the private cloud is chosen as it becomes increasingly important to the organisation. Delivering seamlessly interconnected and ever growing public, private and legacy environments necessitates hosting facilities with fit for purpose networking and plenty of power on tap.