20 October 2016

As UKFast points out, cloud means IT pros have to get used to network topology changing very quickly – but they also have to make rapid changes to their mindsets.

If all the headlines are to be believed, the future for IT professionals lies in a handful of specialist areas where the industry reportedly faces the biggest skills shortage. As well as cyber security and ‘Big Data’, cloud computing is one of the topics that comes up regularly here. 

So if you’ve cut your teeth on LANs and WANs, and now want to run a data centre network that delivers enterprise class cloud services, can you simply use the ‘traditional’ IT skills gained from your corporate experience? Many of the data centre operators and specialist vendors that we spoke to are in no doubt about the answer here.

“In short – no,” says Matt Lovell, CTO of Pulsant. “Cloud exchanges provide the physical interconnection between data centre services and data centres but this is only part of the customer challenge. Many customers seek additional support in terms of forming logical connections between their services and the corresponding cloud service provider services over the physical cloud exchanges.

“In addition, configuring these across multiple geographic or service availability zones in automated failover and recovery topologies is essential for many customers to deliver a similar experience to existing solutions. There is also the need to monitor and manage disparate third-party networks, including wireless and mobile to ensure a consistent user experience.”

Lawrence Jones, CEO of data centre operator UKFast, would agree here, and points out that because clouds are so dense, you get a lot more networking into the same environment. “Previously, you would have had many physical switches and cables within your infrastructure; now there’s the virtual switching layer which sits on top. This requires a different skill set than working solely with steel switches and cables. You still need those traditional skills in your team but now there’s a new top layer of virtualisation which you need to understand so that the two can work together.”

Like many, Jones points out that software defined networking (SDN) is a lot more prevalent now, and while this potentially simplifies the underlying network topology (because it’s all handled through the software) at the same time it adds another layer of complexity where different skills are needed.

Jim Fagan, head of cloud at Australian-based global telecom services operator Telstra, says IT managers need knowledge of SDN to achieve the automation that networking needs in order to “truly” enhance and realise the benefits of cloud.

He adds that capacity management is a completely different challenge when running a cloud network as you must have enough capacity to not only handle spikes in traffic, but also the ability to load balance the network as well.

Understanding and being able to deal with the need for scalability comes up time and again when asking the industry about the skills required for data centres IT manager that are responsible for cloud networks. 

“In our experience, planning for growth and putting contingencies in place so the platform can scale as you need it too is critical,” says Jones. “It’s always better to over-deliver and have more capacity than you need, than to have to run the infrastructure at full tilt. As if that’s not a challenge enough, you also need to plan for something else that’s hard to estimate: an unpredictable workload. Unlike traditional enterprise solutions – where you know where your database servers and your exchange servers are – cloud can be fairly chaotic simply because you’re provisioning the resources and people can use it for whatever they want.”

Jones goes on to say that in cloud environments you tend to see a lot more east-to-west traffic because there’s a huge amount of back-end processing happening. Furthermore, he points out that applications and jobs tend to be rolled out a lot faster in the cloud as virtual environments can be spun up in seconds, placing yet greater demands on resources.

“Cloud pros have to get used to network topology changing very quickly,” says Jones. “As customers provision new resources to their solution they need to ensure that the underlying network is suited to any changes. Because of this, we’re seeing a shift from traditional leaf and spine topologies towards a mesh network topology, giving providers a lot more flexibility in terms of east/west traffic because everything links to everything else within the network.”

According to Jones, all this makes the system a lot more adaptable, but of course it requires the skilled staff to reconfigure and manage the networks accordingly.

Jason Dover, director of product line management at load balancing specialist KEMP Technologies, believes data centre cloud architects need to have the ability to map out the longer-term way that different components in a heterogeneous environment will operate as utilisation scales. He reckons this requires a new mindset that assumes more dynamic change for deployed elements. 

“As an example, on-demand connectivity may need to be established and quickly destroyed for intermittent peak usage bursts.”

Dover goes on to say that an understanding of a cloud operating model, and how the technology being leveraged maps back to business processes, is just as critical as a deep understanding of the WAN connectivity options available and how to architect an IaaS for both high availability and virtual site resiliency. 

Schneider Electric also reckons that mindsets need to be changed. Steven Carlini, the company’s senior director of data centre solutions, IT division, says: “Current thinking on availability focuses on individual sites; future thinking must consider entire hybrid architecture of on premise access to the centralised cloud sites. Skills sets that need to be developed include defining and deploying physical infrastructure (power and cooling) monitoring systems for the remote or branch locations, and understanding which sites are in need of attention or service.”

Michael South, product manager cloud connectivity with Colt Technology Services, is unlikely to disagree here. He says enterprises tend to have some newer “born in the cloud” apps that are typically easy to migrate to the cloud. But he points that most customers also have a legacy to manage, which is less straightforward to manage in a cloud configuration.

“So this is where hybrid solutions come in, linking customer on-site facilities with publicly-hosted capacity via private networking. This delivers the security, performance and reliability which their applications need. It permits the enterprise to manage all their applications in an optimum way. From the network management side this requires a new set of skills, and not just technical. It’s an entirely different business model.”

Equinix offers further explanation here. The US-based company operates over 145 colocation facilities around the world, and says it more than doubled its capacity in Europe with the acquisition of TelecityGroup’s 40+ data centres as part of a $3.8bn deal that was finalised in January. Its UK portfolio includes 12 sites in London and Manchester.

Theo van Andel, Equinix’s, senior manager, field development marketing, says that with the adoption of cloud-based services there is a shift from capex- to opex-based solutions which impacts the total cost of ownership and overall business case. 

“Basically, the business case depends on the actual traffic/cloud utilisation,” says Andel. “As the amount of data varies, the amount of traffic and cloud utilisation will also vary. This results in a cost aspect which is difficult to accurately predict. Hence, constant monitoring is required to adjust the solution in an effort to minimise monthly recurring costs.”

From mindsets to toolsets

So when creating and running a cloud network, what should the data centre network manager look for? And what are the pitfalls to avoid? 

van Andel warns data centre managers against adopting cloud-based services without reassessing their existing network topography. “Without rethinking and probably redesigning the network architecture towards an interconnection oriented architecture, they face challenges of decreased customer experience and satisfaction as well as steeply increased monthly recurring costs.” 

For Pulsant’s Lovell, the challenges are understanding how disparate networks converge to enable connections between services and data streams. But a further dimension for network managers is protecting these services from malicious or disruptive activity. 

“Over the last 12 months, the focus of maximising service availability has required an equivalent if not greater focus on network availability and performance management in comparison to more traditional mechanical and electrical services.”

Lovell says network visualisation tools, which dynamically enable support teams to build application centric outcome views of customer solutions, are key. He points out that as services mature and change over time, the tools can dynamically adapt to capture, record and process the changes to ensure a constantly monitored outcome.

Cloud exchanges provide the physical interconnection between data centre services and data centres, but according to Pulsant this is only part of the customer challenge.

“Documenting live data flows and network connections, with visibility of overall and incremental performance in different parts of networks, is also critical to assist in performance diagnostics and troubleshooting as well as connection optimisation.”

Schneider Electric’s Carlini also believes that when it comes to the mission critical connectivity of remote/branch locations, network availability is the top issue to overcome. 

“The cloud services are in very robust data centres that have built in redundancy from a hardware and application perspective. However, on many occasions, the network access closets are not treated with the same attention, and suffer from lack of redundancy, run time and the necessary cooling.”

Like Lovell, Carlini says monitoring the network becomes more crucial in a data centre cloud environment. He adds that many IT managers actually have the skills to do this – they just need the right tools.

“It’s more important than ever to deploy remote monitoring software with physical security functionality in data centres and network closets connecting to the cloud. If there are many sites, this data can be rolled up into a dashboard.

“There are also data pools being created from many vendors including Schneider. The data collected is sent to cloud pools where it can be used to automate service dispatch. This data can also be put into a larger pool. Analytics can be run to look for trends opportunities to improve performance and prevent issues that could impact availability.”

KEMP’s Dover concurs here. He believes assessment and visibility tools are by far the most critical to provide insight into the readiness of an application to be delivered as a cloud service, or whether a network is properly architected to handle an increased number of services. 

“Without the ability to simulate and make an educated estimation of the quality of experience that end customers will have when consuming services, customer satisfaction and SLAs are likely to suffer.”

He also says that if adopting public cloud into the strategy, maintaining visibility as services and applications span across the data centre firewall is also important, since network architecture by its very design is different in these environments to support multi-tenancy. 

‘One throat to choke’? 

Is the data centre manager more at risk of being the ‘single throat to choke’ in a cloud/outsourced services environment? Does the buck stop with them when it comes to all network problems?

“If properly skilled and with appropriate network planning, the risk for a data centre manager should actually be lowered when managing a cloud services environment,” says Telstra’s Fagan. “Cloud is actually designed to fail, meaning that with proper architecture there should be built-in redundancy to the entire solution, including the network. I would argue this is more resilient than relying on a single point of failure even if it is ‘carrier grade’.”

What has clearly emerged over the last few years is that in order to succeed and even remain relevant today, IT pros should be more business savvy and play a wider role within their organisations. There is therefore an increased need for them to understand business targets and aims. 

“IT used to be siloed but in the cloud/multi-cloud environment this needs to have an integrated approach,” says Equnix’s van Andel. “IT needs to support the business objectives in order to stay competitive.”

Therefore, as opposed to be being more at risk, KEMP advises ‘traditional’ data centre managers to establish cross-functional collaboration across silos, working more closely with newly emerging groups in enterprise IT such as digital performance management and critical lines of business. 

“By adopting this approach, each group is able to contribute more effectively and take shared responsibility,” says Dover. “In much the way as an enterprise adopting public cloud services inherently takes a shared risk between themselves and the cloud service provider, the only way for success within the enterprise is similar thinking and execution model across business units. Key named champions in IT and the business should drive this.”

If anything, the cloud adds more levels of complexity so there are various different areas in which individuals take responsibility. As Jones is keen to point out, it’s a team effort at UKFast and one person never takes the fall. “Ultimately, responsibility for our eCloud lies with the CTO, but responsibility is also devolved across all areas of our networks, infrastructure and service provision.”

It’s a similar situation at Pulsant. Lovell says there is a need for significant expansion in skills and resources in data centre teams to enable a data centre manager to cover all aspects of cloud exchange networks — especially given the competencies and areas of expertise involved, with the primary focus of most data centre managers being operational integrity and efficiencies, as well as maintaining security operations. 

“While the leadership and process governance will be led clearly and effectively by data centre managers to achieve the customer outcome of service availability, this will require investment and expansion of team skills and resources.”

Richard McMahon, head of support and infrastructure at iomart, agrees: “The cloud is meant to be easy but delivering it and managing a network for thousands of servers is not. We’ve invested millions in our people and network to deliver enterprise cloud services as easily as possible, but we’ve done it by defining roles and responsibilities.”

For McMahon, the data centre manager should be the person concentrating on “keeping the lights on”, making sure the cooling is working, and driving the facility’s overall efficiency. But he adds that iomart has multiple data centres and a private network, so it has to keep roles and responsibilities separate. 

“Networks at data centre scale are massively complex creatures and should fall under an operations or infrastructure manager. There has to be a clear separation of skills and responsibility, especially when network skills are so hard to come by. The network might run in the data centre but it’s an entirely different proposition.”

McMahon goes on to state that there are nowhere near enough skilled individuals who truly understand how things operate under the covers. “The DevOps approach doesn’t fix it either – who is going to troubleshoot the underlying infrastructure when the API doesn’t work? Every vendor has a different way of delivering on the scalable network requirement that has its own limits and challenges. Engineers with these skills are really hard to find.”

And perhaps the ultimate challenge that needs to be overcome is to maintain the suitably skilled resources needed to manage increasing complexity when it comes to data centre and cloud networks. 

As Lovell explains, service providers such as Pulsant are increasingly investing in skills development and additional networking expertise to complement existing data centre support teams to deliver customer requirements. 

However, the areas of expertise incorporate a variety of key specialisms, such as: network and application security; network performance management and optimisation; SDN; and centralised policy management. And all that’s on top of competencies and expertise in specific networking standards to interconnect application interfaces between different service provider solutions and standards.

Many commentators, such as van Andel, therefore believe that IT pros will have to gain cloud service provider knowledge in addition to networking and security knowledge. They say hardware specific expertise/certification is becoming less important as the ability to partner is now gaining greater significance.

As Telstra’s Fagan concludes: “Understanding these challenges and how to mitigate them becomes the essential knowledge base. However, the greatest tool a data centre manager can have today is intellectual curiosity and a willingness to learn. Cloud services has created a new paradigm in data centre operations and data centre managers must embrace this change to successfully adapt.”