05 May 2017
With customers demanding increasingly sophisticated network services, how do data centre operators ensure their own in-house networks can keep pace? JAMES HAYES finds out.
Anyone on a tour of a modern data centre will see scant evidence that these processing powerhouses are at the epicentre of the digital transformation that many businesses are going through across the world.
The old indications that something mega was happening behind the scenes have disappeared; the blinking red and green LEDs that used to signal the traffic volumes streaming through are now behind dust-proof panels, and the hum of massed HVAC units has been hushed. Indeed, most data centre interiors are now veritable sanctums of tranquillity – quite a feat given the scale of disruptive forces that these mighty IT estates now enable 24x7.
Andrew Chant, head of networks at Exponential-e, says the volumes of data created by businesses make the data centre one of the most important aspects of commerce.
“The network, by default, has become the aorta of the economy. Data centre network managers now have a lot of responsibility on their shoulders – more so than five years ago. Like the doctor who looks after the head of state, it is their vital job to ensure the underlying health of the economy is in top shape at all time – which brings with it a high level of pressure.”
Data centres themselves are poised to become subject to disruptive change. 2017 is set to see the introduction of multiple technologies that will re-accelerate digital transformation, and at the same time unleash pressures that could cause data centre owners to radically rethink how their facilities are designed, configured and operated.
In addition to the continued exponential growth in web traffic and uptakes in cloud migration, for instance, terrestrial data communications infrastructures will increasingly be utilised by mobile network operators to underpin wireless backhaul, as data loads on their bandwidth intensify.
As is well documented, data is booming. Citing forecasts from IDC, Ciena’s CTO EMEA Joe Marsella says the ‘digital universe’ as measured by storage volumes is set to grow to 44ZB globally by 2020. “Considerable physical – as well as virtual – data centre space needs to be found and managed to accommodate this.
Networks also have to manage increased volumes of customer I/O traffic triggered by the rapid growth of factors [such as] the rise of IoT-connected devices, instant video messaging, and streamed content services.”
Of course, this should certainly mean good business for UK data centre operators. But this surging extra demand brings fresh complexities for data centre managers and planners who need to maintain current capacity while trying to calculate future requirements.
US-based Rocket specialises in software that enables IT professionals and enterprises to run their most critical applications, business processes and data.
John Proctor, the company’s EMEA VP of sales, says: “Workloads are evolving, shifting, and growing at an alarming rate. Similarly, services can equally disappear overnight. Data centre capacity planning is harder than it has ever been, and making capital expenditures often presents huge risk if the workloads do not persist, or never expand to begin with.”
While high-end data centre operators may feel confident that redundant capacity and facility build-outs will empower them to service demand into 2017 and beyond, the escalating demands being made on data centres of whatever tier level spell other challenges on top of capacity provisioning.
Data centres’ internal networked systems – the systems the facility’s engineers rely on to run operations – also have to be up to coping with the booming data volumes and inevitable complexity creep.
Internet of complexity
Independent data centre consultant Alex Rabbetts says that in the past it was okay to be a ‘pure network engineer’ who understood routing protocols such as Cisco IOS or Check Point IPSO.
But nowadays, he says there is greater pressure on network professionals to have a far wider knowledge. “Network professionals need to understand the whole environment from router to UPS, and from Layer 3 switch to chiller unit, because everything is now networked and ‘talking’ to each other.”
While established infrastructure and management solutions such as DCIM, and now SDN and NFV, have been designed for exponential growth and facility systems integration, some industry observers are asking if they are up to coping with hikes in both volumes of traffic and degrees of complexity as new services sear a trail to the data centres’ communications portals.
The question is whether these worthy concepts can still ‘cut it’ for data centres that, some aver, will have to be fundamentally rebuilt if they are to remain fit for purpose.
Most industry watchers concur that it is the IoT network and application rollouts expected this year that could create the biggest demand surge for data centres, especially those in countries where ambitious large-scale pilots (such as the EC’s Horizon 2020 programme) are scheduled.
At the same time, mobile operators and vendors such as Vodafone and Ericsson are racing to introduce their IoT communications products over the top of existing licensed wireless network infrastructure, while solutions providers utilising unlicensed bandwidth options have been building-out separate services.
All parties know that, in contrast to traditional tariffed mobile services that were introduced over a period of years, IoT ventures have to quickly gain massively wide-scale utilisation to secure sound commercial viability.
The impact of a burgeoning IoT on data centres has been a known challenge for at least the last two years. As Gartner research director Fabrizio Biscotti pointed out in 2014: “IoT deployments will generate huge quantities of data that need to be processed and analysed in real-time.”
The number of devices, combined with the volume, velocity and structure of IoT data, will create unprecedented challenges for all aspects of data centre operation.
Biscotti went on to warn that because processing large quantities of IoT data in real-time will increase as a proportion of data centre workloads, data centre managers will face new security, capacity and analytics challenges because real-time business processes will be at stake. As a result, his advice to data centre managers is to deploy more forward-looking capacity management in these areas to be able to proactively meet the business priorities associated with IoT.
Rabbetts suspects that the IoT’s complexity axis may prove a bigger challenge than that of booming traffic volumes. “I don’t really believe that the challenges data centre network managers face will intensify through growth in cloud utilisation or unified communications. But I do think that the challenges of IoT have yet to be fully understood, and that the IoT could bring a whole new set of [complexities] that network professionals will have to get their heads around.”
With offices around the world, Canada-based CENX specialises in service assurance software solutions for mobile, fixed and cloud data centre service providers. The firm’s development head Paul McCluskey suggests a concomitant challenge facing data centre managers in dealing with the diverse set of architectures needed to deliver services.
“There are still traditional applications where users access software and datasets running on servers in the data centre.
However, the data centre software is increasingly talking to a cloud-based PaaS or IaaS, or a user is accessing a cloud-based SaaS application that might be drawing some information (such as by an API) from the enterprise data centre; or an application in one private cloud is talking to another application in another private cloud (perhaps from another service provider) which are both relying upon the enterprise data centre for authorisation, authentication and resources.”
He goes on to suggest that such demands are slowly but surely changing the traditional role of the data centre and, as a result, how it must be managed. This change to the data centre’s ‘job description’ therefore has to be acknowledged and folded into future evolutions.
The need for speed
Whatever reality plays-out over and after the coming 12 months, it nonetheless means that data centre design and configuration will be compelled to change.
Ciena’s Marsella reckons speed of reaction and adoption will become dominating factors. The extended range of service traffic zoomed through data centres will also call for a rethink with respect to how their internal management systems are run. “Data centre managers now need networks that can react seamlessly to spikes in demand, as well as scale efficiently in terms of cost-per-Gbps.”
He reckons SDN has proved itself in this respect: “Through the combination of SDN and next-generation programmable networking equipment, networks can be virtually reconfigured in moments, freeing-up bandwidth for the services where it is needed, as well as cutting the cost of manual reconfiguration of connections.”
Marsella believes that SDN and NFV have become the most highly-anticipated developments in network infrastructure in over a decade. “A software abstraction layer sitting above a set of virtualised hardware resources can transform the way data centres do business – improving service velocity, costs, and business agility.”
For Don Mac Millan, GM of Dimension Data UK’s data centre business, keeping pace with the demands of delivery against the legacy resources available is the top challenge that IT pros face.
He says the speed of requirement has overtaken the speed of deployment of new management capabilities in almost every environment, adding that as emerging technologies continue to be adopted, many network managers will face greater challenges.
“Their existing tools are unlikely to be suitable for these newly-adopted solutions, and it is doubtful they will have the budget to holistically replace these tools.”
As a result, ‘sprawl’ from additional management point products will occur.
Mac Millan goes on to say that there is a drive in-train by most of the larger infrastructure vendors to either consolidate management platforms or to deliver a ‘manager of managers’ suite. “[This] will simplify the operation of environments that have been historically disjointed without it being a ‘science project’ that takes years to return any tangible benefits.”
He adds that because of software-defined data centres, the management context has shifted from a focus on monitoring to one on actual delivery, automation and orchestration, placing the management at the heart of the service delivery capability.
“SDN solutions provide many benefits to an organisation including greater agility with automated operations, improved security through micro-segmentation, potential cost reductions, and providing a mechanism for cross-functional collaboration and orchestration between networking, server, storage, virtualisation and applications teams.”
With that said, is there still a place for legacy DCIM solutions? Mac Millan says yes in certain use cases, but in a reduced capacity where an SDN overlay solution is adopted. “The underlying network infrastructure still needs to be managed but the SDN solution doesn’t have the ability to do so itself.”
Rabbetts agrees that there will soon be a new generation of data centre management tools that will become increasingly important for the data centre manager.
He predicts that these will require greater bandwidth allocated to dedicated management VLANs (which some data centre managers do already), so that the data centre infrastructure can automatically regulate itself within boundaries set by data centre managers: “Soon, the cooling, humidity, power delivery and other aspects of the facility will be able to safe manage themselves via a network; and the savvy data centre manager needs to be ready for that [development].”
DCIM days numbered
Opinion is divided around how effective Data Centre Infrastructure Management platforms will prove as the next iteration of data centre design starts to take shape. DCIM gives managers a range of controls over both the computing and network components, as well as over service delivery and quality assurance, including capacity planning.
The use of DCIM tools is elective, and not all quarters of the industry believe that it offers the best way forward given the welter of challenges such facilities typically will face.
“DCIM has become a very generic term, and DCIM applications now provide a multitude of functionalities – from electrical and mechanical infrastructure management, down to IT optimisation and network management,” says Stefano D’Agostino, software solutions business manager, data centre, Schneider Electric.
He calls for DCIM’s main focus to continue to be based around assets and capacity management for the physical infrastructure in terms of space, power, cooling and networks. “Hence, in my opinion, DCIM on its own is not the right tool to manage and monitor the increasing complexity of SDNs,” says D’Agostino.
Rocket Software’s Proctor reckons DCIM still ‘cuts it’ for basic management requirements. But he warns that if it does not evolve to link and leverage analytics and other business measures, DCIM will be written-out of the picture by application developers collecting data and reporting on them with modern visualisations.
“Data centre managers should be looking to implement dashboard tools such as Geckoboard as a way to consolidate KPIs and data across multiple vendors and technologies. It makes no sense to implement a single vendor solution if you don’t have a solid view of the multiple data sources you use as a business.”
D’Agostino argues that for the optimum operation of both data centre LANs and service networks, a balancing act between DCIM solutions and SDN is key.
“Only complete integration of both DCIM and SDNs will result in 100 per cent network availability and predictive optimisation in case of physical failures or power outages. It is fundamental that DCIM software is correctly deployed, implemented and integrated with SDN software controllers.
"DCIM shall monitor and manage physical assets, while signalling availability risks to predictive network management agents. In this way, the agents can react and reconfigure the network accordingly.”
Rabbetts doesn’t agree. He points out that a management system that is more sophisticated than DCIM is needed, and believes this is not very far away.
“DCIM is really an interim solution. It brings together some functionality in managing the IT infrastructure, along with some functionality from the BMS. But it has never really been a great long-term solution. New management solutions will be intelligent, automated and complex, and when implemented well, will make the data centre manager’s life much easier.”
Who – or what – will run the future data centre?
D’Agostino points out that dedicated tools and standards such as Open Daylight and the Open Network operating system are emerging to manage SDNs.
“These tools allow complete network management abstracted from the physical layer, and provide quick ways of deploying a reconfigured network with the same simplicity of configuring virtual applications. There is also a trend of increased investment into AI engines that will enable companies to relinquish some (or all) of the network management and optimisation tasks that create intelligent and self-learning algorithms.”
Dimension Data adds that there are numerous tools or services that can analyse the workload of a particular asset at any point in time.
Mac Millan says: “The next generation of services for analysis not only interrogates the technical environment, but also maps this to application priority and to cost models to determine how services are being delivered. They also provide an impact analysis which details if technology investments are improving the service delivery.”
In terms of future directions, many data centre insiders look to new thinking in management tools to ensure that managers are able to retain necessary control without inhibiting the data centre’s potential to service additional business opportunities. “Investments take time to return their value, and it’s very easy to invest in something that becomes acquired or irrelevant,” says Mac Millan.
“In the dynamic market that we are experiencing, there will always be the ‘new breed’ of tools that make things simpler to manage. Unfortunately, these solutions often become the victim of [their own] success, and are acquired by a more-established market brand which can, sometimes, slow down their innovation.”
For Rabbetts, the next-generation of data centre network management tools should focus on intelligence rather than management, and take a holistic approach to the running of the data centre and include every aspect of the infrastructure, both IT and mechanical and electrical.
The tools he describes will be capable of knowing, for example, when a server processing load has diminished, and that it is therefore using less power and producing less heat.
“Next-generation intelligent tools will know that the fan belt on an air-conditioning downflow unit needs replacement; that the outside temperature has dropped sufficiently such that free air can be used, requiring only humidification control and filtering. The new tools will be capable of logical decision making based on clearly defined parameters. They will make decisions on behalf of the data centre manager.”